def __init__(dark, config_filepath=None, weight_filepath=None, verbose=VERBOSE_DARK, quiet=QUIET_DARK):
"""
Create the C object for the PyDarknet YOLO detector.
Args:
verbose (bool, optional): verbose flag; defaults to --verbdark flag
Returns:
detector (object): the Darknet YOLO Detector object
"""
dark.CLASS_LIST = None
if config_filepath in ["default", "v2", None]:
config_filepath = ut.grab_file_url(DEFAULT_CONFIG_URL, appname="pydarknet")
dark.CLASS_LIST = CLASS_LIST
elif config_filepath in ["v1", "old", "original"]:
dark.CLASS_LIST = OLD_CLASS_LIST
config_filepath = ut.grab_file_url(OLD_DEFAULT_CONFIG_URL, appname="pydarknet")
if weight_filepath in ["default", "v2", None]:
weight_filepath = ut.grab_file_url(DEFAULT_WEIGHTS_URL, appname="pydarknet")
elif weight_filepath in ["v1", "old", "original"]:
weight_filepath = ut.grab_file_url(OLD_DEFAULT_WEIGHTS_URL, appname="pydarknet")
dark.verbose = verbose
dark.quiet = quiet
dark._load(config_filepath, weight_filepath)
if dark.verbose and not dark.quiet:
print("[pydarknet py] New Darknet_YOLO Object Created")
def models_cnn(
ibs,
config_dict,
parse_classes_func,
parse_line_func,
check_hash=False,
hidden_models=[],
**kwargs,
):
import urllib
model_dict = {}
for config_tag in config_dict:
if config_tag in hidden_models:
continue
try:
config_url = config_dict[config_tag]
classes_url = parse_classes_func(config_url)
try:
classes_filepath = ut.grab_file_url(
classes_url, appname='wbia', check_hash=check_hash
)
assert exists(classes_filepath)
except (urllib.error.HTTPError, AssertionError):
continue
classes_filepath = ut.truepath(classes_filepath)
line_list = parse_line_func(classes_filepath)
model_dict[config_tag] = line_list
except Exception:
pass
return model_dict
def download_tomcat():
"""
Put tomcat into a directory controlled by ibeis
CommandLine:
# Reset
python -c "import utool as ut; ut.delete(ut.unixjoin(ut.get_app_resource_dir('ibeis'), 'tomcat'))"
"""
from os.path import splitext, dirname
print('Grabbing tomcat')
# FIXME: need to make a stable link
if ut.WIN32:
tomcat_binary_url = 'http://mirrors.advancedhosters.com/apache/tomcat/tomcat-8/v8.0.24/bin/apache-tomcat-8.0.24-windows-x86.zip'
else:
tomcat_binary_url = 'http://mirrors.advancedhosters.com/apache/tomcat/tomcat-8/v8.0.24/bin/apache-tomcat-8.0.24.zip'
zip_fpath = ut.grab_file_url(tomcat_binary_url, appname='ibeis')
# Download tomcat into the IBEIS resource directory
tomcat_dpath = join(dirname(zip_fpath), 'tomcat')
if not ut.checkpath(tomcat_dpath, verbose=True):
# hack because unzipping is still weird
ut.unzip_file(zip_fpath)
tomcat_dpath_tmp = splitext(zip_fpath)[0]
ut.move(tomcat_dpath_tmp, tomcat_dpath)
if ut.checkpath(join(tomcat_dpath, 'bin'), verbose=True):
scriptnames = ['catalina.sh', 'startup.sh', 'shutdown.sh']
for fname in scriptnames:
fpath = join(tomcat_dpath, 'bin', fname)
if not ut.is_file_executable(fpath):
print('Adding executable bits to script %r' % (fpath,))
ut.chmod_add_executable(fpath)
return tomcat_dpath
def pretrained_camvid_segnet_basic():
import utool as ut
from pysseg.harness import Harness
from pysseg.tasks import CamVid
workdir = expanduser('~/data/work/camvid/')
arch = 'segnet_basic'
harn = Harness(workdir=workdir, arch=arch)
harn.task = CamVid()
harn.test_batch_size = 1
harn.test_imdir = '~/sseg/SegNet/CamVid/test'
harn.test_gtdir = '~/sseg/SegNet/CamVid/testannot'
harn.gpu_num = gpu_util.find_unused_gpu()
pretrained_weights_fpath = ut.grab_file_url(
'http://mi.eng.cam.ac.uk/~agk34/resources/SegNet/segnet_basic_camvid.caffemodel')
harn.test_weights_fpath = pretrained_weights_fpath
harn.prepare_test_model()
harn.prepare_test_predict_dpath()
"""
diff -u --ignore-space-change \
/data/jon.crall/segnet-exact/SegNet-Tutorial/Models/segnet_basic_inference.prototxt \
/home/local/KHQ/jon.crall/data/work/camvid/models/segnet_basic_predict_hmkcclslvxtlpgwozhiaonfvfmvhigxc.prototext
"""
# harn.prepare_test_input()
harn.evaluate()
def download_tomcat():
"""
Put tomcat into a directory controlled by ibeis
CommandLine:
# Reset
python -c "import utool as ut; ut.delete(ut.unixjoin(ut.get_app_resource_dir('ibeis'), 'tomcat'))"
"""
print('Grabbing tomcat')
# FIXME: need to make a stable link
if ut.WIN32:
tomcat_binary_url = 'http://mirrors.advancedhosters.com/apache/tomcat/tomcat-8/v8.0.36/bin/apache-tomcat-8.0.36-windows-x86.zip'
else:
tomcat_binary_url = 'http://mirrors.advancedhosters.com/apache/tomcat/tomcat-8/v8.0.36/bin/apache-tomcat-8.0.36.zip'
zip_fpath = ut.grab_file_url(tomcat_binary_url, appname='ibeis')
# Download tomcat into the IBEIS resource directory
tomcat_dpath = join(dirname(zip_fpath), 'tomcat')
if not ut.checkpath(tomcat_dpath, verbose=True):
# hack because unzipping is still weird
ut.unzip_file(zip_fpath)
tomcat_dpath_tmp = splitext(zip_fpath)[0]
ut.move(tomcat_dpath_tmp, tomcat_dpath)
if ut.checkpath(join(tomcat_dpath, 'bin'), verbose=True):
scriptnames = ['catalina.sh', 'startup.sh', 'shutdown.sh']
for fname in scriptnames:
fpath = join(tomcat_dpath, 'bin', fname)
if not ut.is_file_executable(fpath):
print('Adding executable bits to script %r' % (fpath, ))
ut.chmod_add_executable(fpath)
return tomcat_dpath
def classify_thumbnail_list(thumbnail_list, model='v1'):
print('[classifier] Loading the classifier training data')
data_list = np.array(thumbnail_list, dtype=np.uint8)
print('[mnist] Loading the data into a JPCNN_Data')
data = JPCNN_Data()
data.set_data_list(data_list)
print('[classifier] Create the JPCNN_Model used for testing')
url = MODEL_DOMAIN + MODEL_URLS[model]
model_path = ut.grab_file_url(url, appname='ibeis')
model = Classifier_Model(model_path)
print('[mnist] Create the JPCNN_network and start testing')
net = JPCNN_Network(model, data)
test_results = net.test('.', best_weights=True)
prediction_list = test_results['label_list']
confidence_list = test_results['confidence_list']
result_list = list(zip(confidence_list, prediction_list))
# Release memory
data = None
model = None
net = None
return result_list
def get_flukebook_image_uuids(ibs):
from datetime import datetime
import uuid
import pytz
PST = pytz.timezone('US/Pacific')
url = 'https://www.flukebook.org/acmIdSync.jsp'
now = datetime.now(tz=PST)
timestamp = now.strftime('%Y-%m-%d-%H-00-00')
filename = 'flukebook.image.admid.%s.json' % (timestamp, )
filepath = ut.grab_file_url(url, appname='wbia', fname=filename)
with open(filepath, 'r') as file:
file_content = file.read()
file_json = ut.from_json(file_content)
logger.info('Loaded %d Image ACM string UUIDs from Flukebook' %
(len(file_json), ))
uuid_list = []
for uuid_str in file_json:
try:
uuid_ = uuid.UUID(uuid_str)
uuid_list.append(uuid_)
except ValueError:
continue
logger.info('Validated %d Image UUIDs from Flukebook' % (len(uuid_list), ))
flukebook_image_uuid_list = list(set(uuid_list))
logger.info('Validated %d de-duplicated Image UUIDs from Flukebook' %
(len(uuid_list), ))
return flukebook_image_uuid_list
def download_image_urls(image_url_info_list):
# Find ones that we already have
print('Requested %d downloaded images' % (len(image_url_info_list)))
full_gpath_list = [join(image_dir, basename(gpath)) for gpath in image_url_info_list]
exists_list = [ut.checkpath(gpath) for gpath in full_gpath_list]
image_url_info_list_ = ut.compress(image_url_info_list, ut.not_list(exists_list))
print('Already have %d/%d downloaded images' % (
len(image_url_info_list) - len(image_url_info_list_), len(image_url_info_list)))
print('Need to download %d images' % (len(image_url_info_list_)))
#import sys
#sys.exit(0)
# Download the rest
imgurl_prefix = 'https://snapshotserengeti.s3.msi.umn.edu/'
image_url_list = [imgurl_prefix + suffix for suffix in image_url_info_list_]
for img_url in ut.ProgressIter(image_url_list, lbl='Downloading image'):
ut.grab_file_url(img_url, download_dir=image_dir)
return full_gpath_list
def label_chip_list(chip_list, model='v1'):
print('[classifier] Loading the classifier training data')
data_list = np.array(chip_list, dtype=np.uint8)
print('[mnist] Loading the data into a JPCNN_Data')
data = JPCNN_Data()
data.set_data_list(data_list)
print('[classifier] Create the JPCNN_Model used for testing')
url = MODEL_DOMAIN + MODEL_URLS[model]
model_path = ut.grab_file_url(url, appname='ibeis')
model = Labeler_Model(model_path)
print('[mnist] Create the JPCNN_network and start testing')
net = JPCNN_Network(model, data)
test_results = net.test('.', best_weights=True)
class_list = list(net.config['data_label_encoder'].classes_)
prediction_list = test_results['label_list']
confidence_list = test_results['confidence_list']
probability_list = test_results['probability_list']
species_list = []
viewpoint_list = []
for prediction in prediction_list:
prediction = prediction.strip()
if ':' in prediction:
prediction = prediction.split(':')
species, viewpoint = prediction
else:
species = prediction
viewpoint = None
if species.lower() == 'ignore':
species = const.UNKNOWN
species_list.append(species)
viewpoint_list.append(viewpoint)
quality_list = [const.QUAL_UNKNOWN] * len(prediction_list)
orientation_list = [0.0] * len(prediction_list)
probability_dict_list = []
for probability in probability_list:
probability_dict = {
class_ : prob
for class_, prob in zip(class_list, probability)
}
probability_dict_list.append(probability_dict)
result_list = list(zip(confidence_list, species_list, viewpoint_list,
quality_list, orientation_list, probability_dict_list))
# Release memory
data = None
model = None
net = None
return result_list
def get_database_icon(ibs, max_dsize=(None, 192), aid=None):
r"""
Args:
max_dsize (tuple): (default = (None, 192))
Returns:
None: None
CommandLine:
python -m wbia.control.IBEISControl --exec-get_database_icon --show
python -m wbia.control.IBEISControl --exec-get_database_icon --show --db Oxford
Example:
>>> # DISABLE_DOCTEST
>>> from wbia.control.IBEISControl import * # NOQA
>>> import wbia
>>> ibs = wbia.opendb(defaultdb='testdb1')
>>> icon = ibs.get_database_icon()
>>> ut.quit_if_noshow()
>>> import wbia.plottool as pt
>>> pt.imshow(icon)
>>> ut.show_if_requested()
"""
# if ibs.get_dbname() == 'Oxford':
# pass
# else:
import vtool as vt
if hasattr(ibs, 'force_icon_aid'):
aid = ibs.force_icon_aid
if aid is None:
species = ibs.get_primary_database_species()
# Use a url to get the icon
url = {
ibs.const.TEST_SPECIES.GIR_MASAI:
'http://i.imgur.com/tGDVaKC.png',
ibs.const.TEST_SPECIES.ZEB_PLAIN:
'http://i.imgur.com/2Ge1PRg.png',
ibs.const.TEST_SPECIES.ZEB_GREVY:
'http://i.imgur.com/PaUT45f.png',
}.get(species, None)
if url is not None:
icon = vt.imread(ut.grab_file_url(url), orient='auto')
else:
# HACK: (this should probably be a db setting)
# use an specific aid to get the icon
aid = {
'Oxford': 73,
'seaturtles': 37
}.get(ibs.get_dbname(), None)
if aid is None:
# otherwise just grab a random aid
aid = ibs.get_valid_aids()[0]
if aid is not None:
icon = ibs.get_annot_chips(aid)
icon = vt.resize_to_maxdims(icon, max_dsize)
return icon
def find_or_download_wilbook_warfile():
r"""
scp [email protected]:/var/lib/tomcat/webapps/ibeis.war \
~/Downloads/pachy_ibeis.war wget
http://dev.wildme.org/ibeis_data_dir/ibeis.war
"""
war_url = 'http://dev.wildme.org/ibeis_data_dir/ibeis.war'
war_fpath = ut.grab_file_url(war_url, appname='ibeis')
return war_fpath
def test(gpath_list, canonical_weight_filepath=None, **kwargs):
from wbia.detecttools.directory import Directory
# Get correct weight if specified with shorthand
archive_url = None
ensemble_index = None
if canonical_weight_filepath is not None and ':' in canonical_weight_filepath:
assert canonical_weight_filepath.count(':') == 1
canonical_weight_filepath, ensemble_index = canonical_weight_filepath.split(
':')
ensemble_index = int(ensemble_index)
if canonical_weight_filepath in ARCHIVE_URL_DICT:
archive_url = ARCHIVE_URL_DICT[canonical_weight_filepath]
archive_path = ut.grab_file_url(archive_url,
appname='wbia',
check_hash=True)
else:
raise RuntimeError('canonical_weight_filepath %r not recognized' %
(canonical_weight_filepath, ))
assert os.path.exists(archive_path)
archive_path = ut.truepath(archive_path)
ensemble_path = archive_path.strip('.zip')
if not os.path.exists(ensemble_path):
ut.unarchive_file(archive_path, output_dir=ensemble_path)
assert os.path.exists(ensemble_path)
direct = Directory(ensemble_path,
include_file_extensions=['weights'],
recursive=True)
weights_path_list = direct.files()
weights_path_list = sorted(weights_path_list)
assert len(weights_path_list) > 0
if ensemble_index is not None:
assert 0 <= ensemble_index and ensemble_index < len(weights_path_list)
weights_path_list = [weights_path_list[ensemble_index]]
assert len(weights_path_list) > 0
logger.info('Using weights in the ensemble: %s ' %
(ut.repr3(weights_path_list), ))
result_list = test_ensemble(gpath_list, weights_path_list, **kwargs)
for result in result_list:
x0 = max(result['x0'], 0.0)
y0 = max(result['y0'], 0.0)
x1 = max(result['x1'], 0.0)
y1 = max(result['y1'], 0.0)
yield (
x0,
y0,
x1,
y1,
)
def get_cls_net(config, **kwargs):
model = HighResolutionNet(config, **kwargs)
model_url = HRNET_PRETRAINED_URL
model_fname = model_url.split('/')[-1]
model_path = ut.grab_file_url(model_url,
appname='wbia_orientation',
check_hash=True,
fname=model_fname)
model.init_weights(model_path)
return model
def ensure_model(model, redownload=False):
try:
url = MODEL_DOMAIN + MODEL_URLS[model]
extracted_fpath = ut.grab_file_url(url, appname='ibeis_cnn',
redownload=redownload,
check_hash=True)
except KeyError as ex:
ut.printex(ex, 'model is not uploaded', iswarning=True)
extracted_fpath = ut.unixjoin(ut.get_app_resource_dir('ibeis_cnn'), model)
ut.assert_exists(extracted_fpath)
return extracted_fpath
def get_database_icon(ibs, max_dsize=(None, 192), aid=None):
r"""
Args:
max_dsize (tuple): (default = (None, 192))
Returns:
None: None
CommandLine:
python -m ibeis.control.IBEISControl --exec-get_database_icon --show
python -m ibeis.control.IBEISControl --exec-get_database_icon --show --db Oxford
Example:
>>> # DISABLE_DOCTEST
>>> from ibeis.control.IBEISControl import * # NOQA
>>> import ibeis
>>> ibs = ibeis.opendb(defaultdb='testdb1')
>>> icon = ibs.get_database_icon()
>>> ut.quit_if_noshow()
>>> import plottool as pt
>>> pt.imshow(icon)
>>> ut.show_if_requested()
"""
#if ibs.get_dbname() == 'Oxford':
# pass
#else:
import vtool as vt
if hasattr(ibs, 'force_icon_aid'):
aid = ibs.force_icon_aid
if aid is None:
species = ibs.get_primary_database_species()
# Use a url to get the icon
url = {
ibs.const.TEST_SPECIES.GIR_MASAI: 'http://i.imgur.com/tGDVaKC.png',
ibs.const.TEST_SPECIES.ZEB_PLAIN: 'http://i.imgur.com/2Ge1PRg.png',
ibs.const.TEST_SPECIES.ZEB_GREVY: 'http://i.imgur.com/PaUT45f.png',
}.get(species, None)
if url is not None:
icon = vt.imread(ut.grab_file_url(url), orient='auto')
else:
# HACK: (this should probably be a db setting)
# use an specific aid to get the icon
aid = {
'Oxford': 73,
'seaturtles': 37,
}.get(ibs.get_dbname(), None)
if aid is None:
# otherwise just grab a random aid
aid = ibs.get_valid_aids()[0]
if aid is not None:
icon = ibs.get_annot_chips(aid)
icon = vt.resize_to_maxdims(icon, max_dsize)
return icon
def _ensure_model_exists(ibs, aid_list, config_path):
species = ibs.get_annot_species_texts(aid_list[0])
model_url = MODEL_URLS[species]
# get expected model location from config file. Couple lines copied from Olga's compute_db.py
with open(config_path) as config_buffer:
config = json.loads(config_buffer.read())
exp_folder = os.path.join(
_PLUGIN_FOLDER, config['train']['exp_dir'], config['train']['exp_id']
)
local_fpath = os.path.join(exp_folder, 'best_weights.h5')
if os.path.isfile(local_fpath):
return True
# download the model and put it in the model_folder
os.makedirs(exp_folder, exist_ok=True)
ut.grab_file_url(model_url, download_dir=exp_folder, fname=local_fpath)
return True
def find_or_download_wilbook_warfile(ensure=True, redownload=False):
r"""
scp [email protected]:/var/lib/tomcat/webapps/ibeis.war \
~/Downloads/pachy_ibeis.war wget
http://dev.wildme.org/ibeis_data_dir/ibeis.war
"""
#war_url = 'http://dev.wildme.org/ibeis_data_dir/ibeis.war'
war_url = 'http://springbreak.wildbook.org/tools/latest.war'
war_fpath = ut.grab_file_url(war_url, appname='ibeis',
ensure=ensure, redownload=redownload,
fname='ibeis.war')
return war_fpath
def setup_kp_network(network_str):
fn = KP_NETWORK_OPTIONS[network_str]['url']
file_url = join('https://lev.cs.rpi.edu/public/models/', fn)
network_params_path = ut.grab_file_url(file_url, appname='ibeis')
network_params = ut.load_cPkl(network_params_path)
# network_params also includes normalization constants needed for the dataset, and is assumed to be a dictionary
# with keys mean, std, and params
network_exp = KP_NETWORK_OPTIONS[network_str]['exp']()
ll.set_all_param_values(network_exp, network_params['params'])
X = T.tensor4()
network_fn = tfn([X], ll.get_output(network_exp, X, deterministic=True))
return {'mean': network_params['mean'], 'std': network_params['std'], 'networkfn': network_fn,
'input_size': KP_NETWORK_OPTIONS[network_str]['size']}
def grab_mnist2():
""" Follows lasange example """
train_data_gz = ut.grab_file_url(
'http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz')
train_labels_gz = ut.grab_file_url(
'http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz')
test_data_gz = ut.grab_file_url(
'http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz')
test_labels_gz = ut.grab_file_url(
'http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz')
train_data = load_mnist_images(train_data_gz)
test_data = load_mnist_images(test_data_gz)
train_labels = load_mnist_labels(train_labels_gz)
test_labels = load_mnist_labels(test_labels_gz)
data = np.vstack((train_data, test_data))
labels = np.append(train_labels, test_labels)
metadata = {}
metadata['splitset'] = ['train'] * len(train_data) + ['test'
] * len(test_labels)
return data, labels, metadata
def find_or_download_wilbook_warfile(ensure=True, redownload=False):
r"""
scp [email protected]:/var/lib/tomcat/webapps/ibeis.war \
~/Downloads/pachy_ibeis.war wget
http://dev.wildme.org/ibeis_data_dir/ibeis.war
"""
#war_url = 'http://dev.wildme.org/ibeis_data_dir/ibeis.war'
war_url = 'http://springbreak.wildbook.org/tools/latest.war'
war_fpath = ut.grab_file_url(war_url,
appname='ibeis',
ensure=ensure,
redownload=redownload,
fname='ibeis.war')
return war_fpath
def install_cockatrice():
cockatrice_url = 'http://www.woogerworks.com/files/cockatrice.weeklybuilds/Cockatrice-WindowsClient.exe'
import utool as ut
fpath = ut.grab_file_url(cockatrice_url)
# run setup script
ut.cmd(fpath)
# press enter a few times
import win32com.client as w32
shell = w32.Dispatch("WScript.Shell")
shell.AppActivate('Cockatrice Setup')
shell.SendKeys("{ENTER}")
shell.SendKeys("{ENTER}")
shell.SendKeys("{ENTER}")
shell.SendKeys("{ENTER}")
def _load_published(self, ibs, species, task_key):
"""
>>> from wbia.algo.verif.vsone import * # NOQA
>>> self = Deployer()
>>> species = 'zebra_plains'
>>> task_key = 'match_state'
"""
base_url = 'https://{remote}/public/models/pairclf'.format(
**self.publish_info)
task_fnames = self.published[species]
fname = task_fnames[task_key]
grabkw = dict(appname='wbia', check_hash=False, verbose=0)
meta_url = base_url + '/' + fname + self.meta_suffix
meta_fpath = ut.grab_file_url(meta_url, **grabkw) # NOQA
deploy_url = base_url + '/' + fname
deploy_fpath = ut.grab_file_url(deploy_url, **grabkw)
verif = self._make_verifier(ibs, deploy_fpath, task_key)
return verif
def setup_te_network(network_str):
fn = TE_NETWORK_OPTIONS[network_str]['url']
file_url = join('https://lev.cs.rpi.edu/public/models/', fn)
network_params_path = ut.grab_file_url(file_url, appname='ibeis')
network_params = ut.load_cPkl(network_params_path)
# network_params also includes normalization constants needed for the dataset, and is assumed to be a dictionary
# with keys mean, std, and params
network_exp = TE_NETWORK_OPTIONS[network_str]['exp']()
ll.set_all_param_values(network_exp, network_params['params'])
X = T.tensor4()
network_fn = tfn([X], ll.get_output(
network_exp[-1], X, deterministic=True))
retdict = {'mean': network_params['mean'], 'std': network_params[
'std'], 'networkfn': network_fn}
if any([i in network_str for i in ('upsample', 'jet')]):
retdict['mod_acc'] = 8
return retdict
def setup_kp_network(network_str):
fn = KP_NETWORK_OPTIONS[network_str]['url']
file_url = join('https://lev.cs.rpi.edu/public/models/', fn)
network_params_path = ut.grab_file_url(file_url, appname='ibeis')
network_params = ut.load_cPkl(network_params_path)
# network_params also includes normalization constants needed for the dataset, and is assumed to be a dictionary
# with keys mean, std, and params
network_exp = KP_NETWORK_OPTIONS[network_str]['exp']()
ll.set_all_param_values(network_exp, network_params['params'])
X = T.tensor4()
network_fn = tfn([X], ll.get_output(network_exp, X, deterministic=True))
return {
'mean': network_params['mean'],
'std': network_params['std'],
'networkfn': network_fn,
'input_size': KP_NETWORK_OPTIONS[network_str]['size']
}
def test_pygist():
print('[pygist] Testing pygist')
# Ensure you have test data
print('[pygist] Ensuring testdata')
datafile = utool.grab_file_url(TEST_MODEL_URL, appname='utool')
test_image_dir = utool.grab_zipped_url(TEST_IMAGES_URL, appname='utool')
imgpaths = utool.list_images(test_image_dir, fullpath=True) # test image paths
outdir = utool.get_app_resource_dir('pygist') # where to put results
# Run pygist on test images
print('[pygist] Running tests')
test_results = pygist.test(imgpaths, outdir=outdir, datafile=datafile)
# Print results
target_results = [-1, -1, 1, -1, 1, -1, -1, -1, 1, 1, -1, 1, 1]
assert target_results == target_results, 'results do not match'
print('test_results = %r' % (test_results,))
print(utool.list_str(list(izip(imgpaths, test_results))))
return locals()
def test_pygist():
print('[pygist] Testing pygist')
# Ensure you have test data
print('[pygist] Ensuring testdata')
datafile = utool.grab_file_url(TEST_MODEL_URL, appname='utool')
test_image_dir = utool.grab_zipped_url(TEST_IMAGES_URL, appname='utool')
imgpaths = utool.list_images(test_image_dir,
fullpath=True) # test image paths
outdir = utool.get_app_resource_dir('pygist') # where to put results
# Run pygist on test images
print('[pygist] Running tests')
test_results = pygist.test(imgpaths, outdir=outdir, datafile=datafile)
# Print results
target_results = [-1, -1, 1, -1, 1, -1, -1, -1, 1, 1, -1, 1, 1]
assert target_results == target_results, 'results do not match'
print('test_results = %r' % (test_results, ))
print(utool.list_str(list(izip(imgpaths, test_results))))
return locals()
def grab_example_smart_xml_fpath():
""" Gets smart example xml
CommandLine:
python -m ibeis.init.sysres --test-grab_example_smart_xml_fpath
Example:
>>> # DISABLE_DOCTEST
>>> import ibeis
>>> import os
>>> smart_xml_fpath = ibeis.sysres.grab_example_smart_xml_fpath()
>>> os.system('gvim ' + smart_xml_fpath)
>>> #ut.editfile(smart_xml_fpath)
"""
smart_xml_url = 'https://www.dropbox.com/s/g1mpjzp57wfnhk6/LWC_000261.xml'
smart_sml_fpath = ut.grab_file_url(smart_xml_url, ensure=True, appname='ibeis')
return smart_sml_fpath
def setup_te_network(network_str):
fn = TE_NETWORK_OPTIONS[network_str]['url']
file_url = join('https://lev.cs.rpi.edu/public/models/', fn)
network_params_path = ut.grab_file_url(file_url, appname='ibeis')
network_params = ut.load_cPkl(network_params_path)
# network_params also includes normalization constants needed for the dataset, and is assumed to be a dictionary
# with keys mean, std, and params
network_exp = TE_NETWORK_OPTIONS[network_str]['exp']()
ll.set_all_param_values(network_exp, network_params['params'])
X = T.tensor4()
network_fn = tfn([X], ll.get_output(network_exp[-1], X,
deterministic=True))
retdict = {
'mean': network_params['mean'],
'std': network_params['std'],
'networkfn': network_fn
}
if any([i in network_str for i in ('upsample', 'jet')]):
retdict['mod_acc'] = 8
return retdict
def grab_example_smart_xml_fpath():
""" Gets smart example xml
CommandLine:
python -m ibeis.init.sysres --test-grab_example_smart_xml_fpath
Example:
>>> # DISABLE_DOCTEST
>>> import ibeis
>>> import os
>>> smart_xml_fpath = ibeis.sysres.grab_example_smart_xml_fpath()
>>> os.system('gvim ' + smart_xml_fpath)
>>> #ut.editfile(smart_xml_fpath)
"""
smart_xml_url = 'https://lev.cs.rpi.edu/public/data/LWC_000261.xml'
smart_sml_fpath = ut.grab_file_url(smart_xml_url,
ensure=True,
appname='ibeis')
return smart_sml_fpath
def load_assigner_classifier(ibs, aid_list, fallback_species='wild_dog'):
species_with_part = ibs.get_annot_species(aid_list[0])
species = species_with_part.split('+')[0]
if species in INMEM_ASSIGNER_MODELS.keys():
clf = INMEM_ASSIGNER_MODELS[species]
else:
if species not in SPECIES_CONFIG_MAP.keys():
print(
'WARNING: Assigner called for species %s which does not have an assigner modelfile specified. Falling back to the model for %s'
% species,
fallback_species,
)
species = fallback_species
model_url = SPECIES_CONFIG_MAP[species]['model_url']
model_fpath = ut.grab_file_url(model_url)
from joblib import load
clf = load(model_fpath)
return clf
def pretrained_camvid_segnet_driving():
import utool as ut
from pysseg.harness import Harness
from pysseg.tasks import CamVid
workdir = expanduser('~/data/work/camvid/')
arch = 'segnet_proper' # not really
harn = Harness(workdir=workdir, arch=arch)
harn.task = CamVid()
harn.test_imdir = '~/sseg/SegNet/CamVid/test'
harn.test_gtdir = '~/sseg/SegNet/CamVid/testannot'
harn.test_suffix = 'pretrained'
harn.gpu_num = gpu_util.find_unused_gpu()
pretrained_weights_fpath = ut.grab_file_url(
'http://mi.eng.cam.ac.uk/~agk34/resources/SegNet/segnet_weights_driving_webdemo.caffemodel')
harn.test_weights_fpath = pretrained_weights_fpath
harn.prepare_test_model()
# harn.prepare_test_input()
# Driving demo seems to have a different model...
# perhaps the labels were trained differently here
# I THINK THIS ONE WAS TRAINED WITH AN EXTRA CATEGORY
harn.test_model_fpath = expanduser('~/sseg/SegNet/')
harn.evaluate()
def detect(gpath_list, config_filepath, weight_filepath, class_filepath, sensitivity,
verbose=VERBOSE_SS, use_gpu=True, use_gpu_id=0,
**kwargs):
"""
Args:
gpath_list (list of str): the list of image paths that need proposal candidates
Kwargs (optional): refer to the SSD documentation for configuration settings
Returns:
iter
"""
def _get_label_name(class_labelmap, label_list):
if not isinstance(label_list, list):
label_list = [label_list]
item_list = class_labelmap.item
name_list = []
for label in label_list:
found = False
for i in range(len(item_list)):
if label == item_list[i].label:
found = True
name_list.append(item_list[i].display_name)
break
assert found
return name_list
# Get correct config if specified with shorthand
config_url = None
if config_filepath in CONFIG_URL_DICT:
config_url = CONFIG_URL_DICT[config_filepath]
config_filepath = ut.grab_file_url(config_url, appname='ibeis',
check_hash=True)
# Get correct weights if specified with shorthand
if weight_filepath in CONFIG_URL_DICT:
if weight_filepath is None and config_url is not None:
config_url_ = config_url
else:
config_url_ = CONFIG_URL_DICT[weight_filepath]
weight_url = _parse_weight_from_cfg(config_url_)
weight_filepath = ut.grab_file_url(weight_url, appname='ibeis',
check_hash=True)
if class_filepath is None:
class_url = _parse_classes_from_cfg(config_url)
class_filepath = ut.grab_file_url(class_url, appname='ibeis',
check_hash=True, verbose=verbose)
# load class labels
with open(class_filepath, 'r') as class_file:
class_labelmap = caffe_pb2.LabelMap()
class_str = str(file.read(class_file))
text_format.Merge(class_str, class_labelmap)
# Need to convert unicode strings to Python strings to support Boost Python
# call signatures in caffe
prototxt_filepath = str(config_filepath) # alias to Caffe nomenclature
caffemodel_filepath = str(weight_filepath) # alias to Caffe nomenclature
assert exists(prototxt_filepath), 'Specified prototxt file not found'
assert exists(caffemodel_filepath), 'Specified caffemodel file not found'
if use_gpu:
caffe.set_mode_gpu()
caffe.set_device(use_gpu_id)
else:
caffe.set_mode_cpu()
net = caffe.Net(prototxt_filepath, caffemodel_filepath, caffe.TEST)
# Determine input size from prototext
with open(prototxt_filepath, 'r') as prototxt_file:
# load all lines
line_list = prototxt_file.readlines()
# look for dim size lines
line_list = [line for line in line_list if 'dim:' in line]
line_list = line_list[:4]
# Get last line
line = line_list[-1]
line_ = line.strip().split(' ')
# Filter empty spaces
line_ = [ _ for _ in line_ if len(_) > 0]
# Get last value on line, which should be the image size
image_resize = int(line_[-1])
# Check to make sure
assert image_resize in [300, 500, 512]
print('FOUND image_resize = %r' % (image_resize, ))
# Input preprocessing: 'data' is the name of the input blob == net.inputs[0]
transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})
transformer.set_transpose('data', (2, 0, 1))
# Mean pixel value
transformer.set_mean('data', np.array([104, 117, 123]))
# The reference model operates on images in [0,255] range instead of [0,1]
transformer.set_raw_scale('data', 255)
# The reference model has channels in BGR order instead of RGB
transformer.set_channel_swap('data', (2, 1, 0))
# Set batch size to 1 and set testing image size
net.blobs['data'].reshape(1, 3, image_resize, image_resize)
results_list_ = []
for gpath in gpath_list:
image = caffe.io.load_image(gpath)
transformed_image = transformer.preprocess('data', image)
net.blobs['data'].data[...] = transformed_image
# Forward pass.
detections = net.forward()['detection_out']
# Parse the outputs.
det_label = detections[0, 0, : , 1]
det_conf = detections[0, 0, : , 2]
det_xmin = detections[0, 0, : , 3]
det_ymin = detections[0, 0, : , 4]
det_xmax = detections[0, 0, : , 5]
det_ymax = detections[0, 0, : , 6]
# Get detections with confidence higher than 0.6.
top_indices = [
i for
i, conf in enumerate(det_conf)
if conf >= sensitivity
]
top_conf = det_conf[top_indices]
top_label_indices = det_label[top_indices].tolist()
top_labels = _get_label_name(class_labelmap, top_label_indices)
top_xmin = det_xmin[top_indices]
top_ymin = det_ymin[top_indices]
top_xmax = det_xmax[top_indices]
top_ymax = det_ymax[top_indices]
height, width = image.shape[:2]
# Compile results
result_list_ = []
zipped = zip(top_xmin, top_ymin, top_xmax, top_ymax, top_labels, top_conf)
for (xmin, ymin, xmax, ymax, label, conf) in zipped:
xtl = int(np.around(xmin * width))
ytl = int(np.around(ymin * height))
xbr = int(np.around(xmax * width))
ybr = int(np.around(ymax * height))
confidence = float(conf)
result_dict = {
'xtl' : xtl,
'ytl' : ytl,
'width' : xbr - xtl,
'height' : ybr - ytl,
'class' : label,
'confidence' : confidence,
}
result_list_.append(result_dict)
results_list_.append(result_list_)
results_list = zip(gpath_list, results_list_)
return results_list
def detect(gpath_list,
config_filepath,
weight_filepath,
class_filepath,
sensitivity,
verbose=VERBOSE_SS,
use_gpu=True,
use_gpu_id=0,
**kwargs):
"""
Args:
gpath_list (list of str): the list of image paths that need proposal candidates
Kwargs (optional): refer to the Faster R-CNN documentation for configuration settings
Returns:
iter
"""
cfg.TEST.HAS_RPN = True # Use RPN for proposals
# Get correct config if specified with shorthand
config_url = None
if config_filepath in CONFIG_URL_DICT:
config_url = CONFIG_URL_DICT[config_filepath]
config_filepath = ut.grab_file_url(config_url,
appname='ibeis',
check_hash=True)
# Get correct weights if specified with shorthand
if weight_filepath in CONFIG_URL_DICT:
if weight_filepath is None and config_url is not None:
config_url_ = config_url
else:
config_url_ = CONFIG_URL_DICT[weight_filepath]
weight_url = _parse_weight_from_cfg(config_url_)
weight_filepath = ut.grab_file_url(weight_url,
appname='ibeis',
check_hash=True)
if class_filepath is None:
class_url = _parse_classes_from_cfg(config_url)
class_filepath = ut.grab_file_url(class_url,
appname='ibeis',
check_hash=True,
verbose=verbose)
class_list = _parse_class_list(class_filepath)
# Need to convert unicode strings to Python strings to support Boost Python
# call signatures in caffe
prototxt_filepath = str(config_filepath) # alias to Caffe nomenclature
caffemodel_filepath = str(weight_filepath) # alias to Caffe nomenclature
assert exists(prototxt_filepath), 'Specified prototxt file not found'
assert exists(caffemodel_filepath), 'Specified caffemodel file not found'
if use_gpu:
caffe.set_mode_gpu()
caffe.set_device(use_gpu_id)
cfg.GPU_ID = use_gpu_id
else:
caffe.set_mode_cpu()
net = caffe.Net(prototxt_filepath, caffemodel_filepath, caffe.TEST)
# Warm-up network on a dummy image
im = 128 * np.ones((300, 500, 3), dtype=np.uint8)
for i in range(2):
_, _ = im_detect(net, im)
results_list_ = []
for gpath in gpath_list:
image = cv2.imread(gpath)
score_list, bbox_list = im_detect(net, image)
# Compile results
result_list_ = []
for class_index, class_name in enumerate(class_list[1:]):
class_index += 1 # because we skipped background
class_boxes = bbox_list[:, 4 * class_index:4 * (class_index + 1)]
class_scores = score_list[:, class_index]
dets_list = np.hstack((class_boxes, class_scores[:, np.newaxis]))
dets_list = dets_list.astype(np.float32)
# # Perform NMS
# keep_list = nms(dets_list, nms_sensitivity)
# dets_list = dets_list[keep_list, :]
# Perform sensitivity check
keep_list = np.where(dets_list[:, -1] >= sensitivity)[0]
dets_list = dets_list[keep_list, :]
for (xtl, ytl, xbr, ybr, conf) in dets_list:
xtl = int(np.around(xtl))
ytl = int(np.around(ytl))
xbr = int(np.around(xbr))
ybr = int(np.around(ybr))
confidence = float(conf)
result_dict = {
'xtl': xtl,
'ytl': ytl,
'width': xbr - xtl,
'height': ybr - ytl,
'class': class_name,
'confidence': confidence,
}
result_list_.append(result_dict)
results_list_.append(result_list_)
results_list = zip(gpath_list, results_list_)
return results_list
def classify(vector_list, weight_filepath, verbose=VERBOSE_SVM, **kwargs):
"""
Args:
thumbail_list (list of str): the list of image thumbnails that need classifying
Returns:
iter
"""
import multiprocessing
import numpy as np
# Get correct weight if specified with shorthand
if weight_filepath in CONFIG_URL_DICT:
weight_url = CONFIG_URL_DICT[weight_filepath]
if weight_url.endswith('.zip'):
weight_filepath = ut.grab_zipped_url(weight_url, appname='ibeis')
else:
weight_filepath = ut.grab_file_url(weight_url, appname='ibeis',
check_hash=True)
# Get ensemble
is_ensemble = isdir(weight_filepath)
if is_ensemble:
weight_filepath_list = sorted([
join(weight_filepath, filename) for filename in listdir(weight_filepath)
if isfile(join(weight_filepath, filename))
])
else:
weight_filepath_list = [weight_filepath]
num_weights = len(weight_filepath_list)
assert num_weights > 0
# Form dictionaries
num_vectors = len(vector_list)
index_list = list(range(num_vectors))
# Generate parallelized wrapper
OLD = False
if is_ensemble and OLD:
vectors_list = [ vector_list for _ in range(num_weights) ]
args_list = zip(weight_filepath_list, vectors_list)
nTasks = num_weights
print('Processing ensembles in parallel using %d ensembles' % (num_weights, ))
else:
num_cpus = multiprocessing.cpu_count()
vector_batch = int(np.ceil(float(num_vectors) / num_cpus))
vector_rounds = int(np.ceil(float(num_vectors) / vector_batch))
args_list = []
for vector_round in range(vector_rounds):
start_index = vector_round * vector_batch
stop_index = (vector_round + 1) * vector_batch
assert start_index < num_vectors
stop_index = min(stop_index, num_vectors)
# print('Slicing index range: [%r, %r)' % (start_index, stop_index, ))
# Slice gids and get feature data
index_list_ = list(range(start_index, stop_index))
vector_list_ = vector_list[start_index: stop_index]
assert len(index_list_) == len(vector_list_)
for weight_filepath in weight_filepath_list:
args = (weight_filepath, vector_list_, index_list_)
args_list.append(args)
nTasks = len(args_list)
print('Processing vectors in parallel using vector_batch = %r' % (vector_batch, ))
# Perform inference
classify_iter = ut.generate2(classify_helper, args_list, nTasks=nTasks,
ordered=True, force_serial=False)
# Classify with SVM for each image vector
score_dict = { index: [] for index in index_list }
class_dict = { index: [] for index in index_list }
for score_dict_, class_dict_ in classify_iter:
for index in index_list:
if index in score_dict_:
score_dict[index] += score_dict_[index]
if index in class_dict_:
class_dict[index] += class_dict_[index]
# Organize and compute mode and average for class and score
for index in index_list:
score_list_ = score_dict[index]
class_list_ = class_dict[index]
score_ = sum(score_list_) / len(score_list_)
class_ = max(set(class_list_), key=class_list_.count)
class_ = 'positive' if int(class_) == 1 else 'negative'
yield score_, class_
def get_injured_sharks():
"""
>>> from wbia.scripts.getshark import * # NOQA
"""
import requests
url = 'http://www.whaleshark.org/getKeywordImages.jsp'
resp = requests.get(url)
assert resp.status_code == 200
keywords = resp.json()['keywords']
key_list = ut.take_column(keywords, 'indexName')
key_to_nice = {k['indexName']: k['readableName'] for k in keywords}
injury_patterns = [
'injury',
'net',
'hook',
'trunc',
'damage',
'scar',
'nicks',
'bite',
]
injury_keys = [
key for key in key_list if any([pat in key for pat in injury_patterns])
]
noninjury_keys = ut.setdiff(key_list, injury_keys)
injury_nice = ut.lmap(lambda k: key_to_nice[k], injury_keys) # NOQA
noninjury_nice = ut.lmap(lambda k: key_to_nice[k], noninjury_keys) # NOQA
key_list = injury_keys
keyed_images = {}
for key in ut.ProgIter(key_list, lbl='reading index', bs=True):
key_url = url + '?indexName={indexName}'.format(indexName=key)
key_resp = requests.get(key_url)
assert key_resp.status_code == 200
key_imgs = key_resp.json()['images']
keyed_images[key] = key_imgs
key_hist = {key: len(imgs) for key, imgs in keyed_images.items()}
key_hist = ut.sort_dict(key_hist, 'vals')
logger.info(ut.repr3(key_hist))
nice_key_hist = ut.map_dict_keys(lambda k: key_to_nice[k], key_hist)
nice_key_hist = ut.sort_dict(nice_key_hist, 'vals')
logger.info(ut.repr3(nice_key_hist))
key_to_urls = {
key: ut.take_column(vals, 'url')
for key, vals in keyed_images.items()
}
overlaps = {}
import itertools
overlap_img_list = []
for k1, k2 in itertools.combinations(key_to_urls.keys(), 2):
overlap_imgs = ut.isect(key_to_urls[k1], key_to_urls[k2])
num_overlap = len(overlap_imgs)
overlaps[(k1, k2)] = num_overlap
overlaps[(k1, k1)] = len(key_to_urls[k1])
if num_overlap > 0:
# logger.info('[%s][%s], overlap=%r' % (k1, k2, num_overlap))
overlap_img_list.extend(overlap_imgs)
all_img_urls = list(set(ut.flatten(key_to_urls.values())))
num_all = len(all_img_urls) # NOQA
logger.info('num_all = %r' % (num_all, ))
# Determine super-categories
categories = ['nicks', 'scar', 'trunc']
# Force these keys into these categories
key_to_cat = {'scarbite': 'other_injury'}
cat_to_keys = ut.ddict(list)
for key in key_to_urls.keys():
flag = 1
if key in key_to_cat:
cat = key_to_cat[key]
cat_to_keys[cat].append(key)
continue
for cat in categories:
if cat in key:
cat_to_keys[cat].append(key)
flag = 0
if flag:
cat = 'other_injury'
cat_to_keys[cat].append(key)
cat_urls = ut.ddict(list)
for cat, keys in cat_to_keys.items():
for key in keys:
cat_urls[cat].extend(key_to_urls[key])
cat_hist = {}
for cat in list(cat_urls.keys()):
cat_urls[cat] = list(set(cat_urls[cat]))
cat_hist[cat] = len(cat_urls[cat])
logger.info(ut.repr3(cat_to_keys))
logger.info(ut.repr3(cat_hist))
key_to_cat = dict([(val, key) for key, vals in cat_to_keys.items()
for val in vals])
# ingestset = {
# '__class__': 'ImageSet',
# 'images': ut.ddict(dict)
# }
# for key, key_imgs in keyed_images.items():
# for imgdict in key_imgs:
# url = imgdict['url']
# encid = imgdict['correspondingEncounterNumber']
# # Make structure
# encdict = encounters[encid]
# encdict['__class__'] = 'Encounter'
# imgdict = ut.delete_keys(imgdict.copy(), ['correspondingEncounterNumber'])
# imgdict['__class__'] = 'Image'
# cat = key_to_cat[key]
# annotdict = {'relative_bbox': [.01, .01, .98, .98], 'tags': [cat, key]}
# annotdict['__class__'] = 'Annotation'
# # Ensure structures exist
# encdict['images'] = encdict.get('images', [])
# imgdict['annots'] = imgdict.get('annots', [])
# # Add an image to this encounter
# encdict['images'].append(imgdict)
# # Add an annotation to this image
# imgdict['annots'].append(annotdict)
# # http://springbreak.wildbook.org/rest/org.ecocean.Encounter/1111
# get_enc_url = 'http://www.whaleshark.org/rest/org.ecocean.Encounter/%s' % (encid,)
# resp = requests.get(get_enc_url)
# logger.info(ut.repr3(encdict))
# logger.info(ut.repr3(encounters))
# Download the files to the local disk
# fpath_list =
all_urls = ut.unique(
ut.take_column(
ut.flatten(
ut.dict_subset(keyed_images,
ut.flatten(cat_to_keys.values())).values()),
'url',
))
dldir = ut.truepath('~/tmpsharks')
from os.path import commonprefix, basename # NOQA
prefix = commonprefix(all_urls)
suffix_list = [url_[len(prefix):] for url_ in all_urls]
fname_list = [suffix.replace('/', '--') for suffix in suffix_list]
fpath_list = []
for url, fname in ut.ProgIter(zip(all_urls, fname_list),
lbl='downloading imgs',
freq=1):
fpath = ut.grab_file_url(url,
download_dir=dldir,
fname=fname,
verbose=False)
fpath_list.append(fpath)
# Make sure we keep orig info
# url_to_keys = ut.ddict(list)
url_to_info = ut.ddict(dict)
for key, imgdict_list in keyed_images.items():
for imgdict in imgdict_list:
url = imgdict['url']
info = url_to_info[url]
for k, v in imgdict.items():
info[k] = info.get(k, [])
info[k].append(v)
info['keys'] = info.get('keys', [])
info['keys'].append(key)
# url_to_keys[url].append(key)
info_list = ut.take(url_to_info, all_urls)
for info in info_list:
if len(set(info['correspondingEncounterNumber'])) > 1:
assert False, 'url with two different encounter nums'
# Combine duplicate tags
hashid_list = [
ut.get_file_uuid(fpath_, stride=8)
for fpath_ in ut.ProgIter(fpath_list, bs=True)
]
groupxs = ut.group_indices(hashid_list)[1]
# Group properties by duplicate images
# groupxs = [g for g in groupxs if len(g) > 1]
fpath_list_ = ut.take_column(ut.apply_grouping(fpath_list, groupxs), 0)
url_list_ = ut.take_column(ut.apply_grouping(all_urls, groupxs), 0)
info_list_ = [
ut.map_dict_vals(ut.flatten, ut.dict_accum(*info_))
for info_ in ut.apply_grouping(info_list, groupxs)
]
encid_list_ = [
ut.unique(info_['correspondingEncounterNumber'])[0]
for info_ in info_list_
]
keys_list_ = [ut.unique(info_['keys']) for info_ in info_list_]
cats_list_ = [ut.unique(ut.take(key_to_cat, keys)) for keys in keys_list_]
clist = ut.ColumnLists({
'gpath': fpath_list_,
'url': url_list_,
'encid': encid_list_,
'key': keys_list_,
'cat': cats_list_,
})
# for info_ in ut.apply_grouping(info_list, groupxs):
# info = ut.dict_accum(*info_)
# info = ut.map_dict_vals(ut.flatten, info)
# x = ut.unique(ut.flatten(ut.dict_accum(*info_)['correspondingEncounterNumber']))
# if len(x) > 1:
# info = info.copy()
# del info['keys']
# logger.info(ut.repr3(info))
flags = ut.lmap(ut.fpath_has_imgext, clist['gpath'])
clist = clist.compress(flags)
import wbia
ibs = wbia.opendb('WS_Injury', allow_newdir=True)
gid_list = ibs.add_images(clist['gpath'])
clist['gid'] = gid_list
failed_flags = ut.flag_None_items(clist['gid'])
logger.info('# failed %s' % (sum(failed_flags), ))
passed_flags = ut.not_list(failed_flags)
clist = clist.compress(passed_flags)
ut.assert_all_not_None(clist['gid'])
# ibs.get_image_uris_original(clist['gid'])
ibs.set_image_uris_original(clist['gid'], clist['url'], overwrite=True)
# ut.zipflat(clist['cat'], clist['key'])
if False:
# Can run detection instead
clist['tags'] = ut.zipflat(clist['cat'])
aid_list = ibs.use_images_as_annotations(clist['gid'],
adjust_percent=0.01,
tags_list=clist['tags'])
aid_list
import wbia.plottool as pt
from wbia import core_annots
pt.qt4ensure()
# annots = ibs.annots()
# aids = [1, 2]
# ibs.depc_annot.get('hog', aids , 'hog')
# ibs.depc_annot.get('chip', aids, 'img')
for aid in ut.InteractiveIter(ibs.get_valid_aids()):
hogs = ibs.depc_annot.d.get_hog_hog([aid])
chips = ibs.depc_annot.d.get_chips_img([aid])
chip = chips[0]
hogimg = core_annots.make_hog_block_image(hogs[0])
pt.clf()
pt.imshow(hogimg, pnum=(1, 2, 1))
pt.imshow(chip, pnum=(1, 2, 2))
fig = pt.gcf()
fig.show()
fig.canvas.draw()
# logger.info(len(groupxs))
# if False:
# groupxs = ut.find_duplicate_items(ut.lmap(basename, suffix_list)).values()
# logger.info(ut.repr3(ut.apply_grouping(all_urls, groupxs)))
# # FIX
# for fpath, fname in zip(fpath_list, fname_list):
# if ut.checkpath(fpath):
# ut.move(fpath, join(dirname(fpath), fname))
# logger.info('fpath = %r' % (fpath,))
# import wbia
# from wbia.dbio import ingest_dataset
# dbdir = wbia.sysres.lookup_dbdir('WS_ALL')
# self = ingest_dataset.Ingestable2(dbdir)
if False:
# Show overlap matrix
import wbia.plottool as pt
import pandas as pd
import numpy as np
dict_ = overlaps
s = pd.Series(dict_, index=pd.MultiIndex.from_tuples(overlaps))
df = s.unstack()
lhs, rhs = df.align(df.T)
df = lhs.add(rhs, fill_value=0).fillna(0)
label_texts = df.columns.values
def label_ticks(label_texts):
import wbia.plottool as pt
truncated_labels = [repr(lbl[0:100]) for lbl in label_texts]
ax = pt.gca()
ax.set_xticks(list(range(len(label_texts))))
ax.set_xticklabels(truncated_labels)
[lbl.set_rotation(-55) for lbl in ax.get_xticklabels()]
[
lbl.set_horizontalalignment('left')
for lbl in ax.get_xticklabels()
]
# xgrid, ygrid = np.meshgrid(range(len(label_texts)), range(len(label_texts)))
# pt.plot_surface3d(xgrid, ygrid, disjoint_mat)
ax.set_yticks(list(range(len(label_texts))))
ax.set_yticklabels(truncated_labels)
[
lbl.set_horizontalalignment('right')
for lbl in ax.get_yticklabels()
]
[
lbl.set_verticalalignment('center')
for lbl in ax.get_yticklabels()
]
# [lbl.set_rotation(20) for lbl in ax.get_yticklabels()]
# df = df.sort(axis=0)
# df = df.sort(axis=1)
sortx = np.argsort(df.sum(axis=1).values)[::-1]
df = df.take(sortx, axis=0)
df = df.take(sortx, axis=1)
fig = pt.figure(fnum=1)
fig.clf()
mat = df.values.astype(np.int32)
mat[np.diag_indices(len(mat))] = 0
vmax = mat[(1 - np.eye(len(mat))).astype(np.bool)].max()
import matplotlib.colors
norm = matplotlib.colors.Normalize(vmin=0, vmax=vmax, clip=True)
pt.plt.imshow(mat, cmap='hot', norm=norm, interpolation='none')
pt.plt.colorbar()
pt.plt.grid('off')
label_ticks(label_texts)
fig.tight_layout()
# overlap_df = pd.DataFrame.from_dict(overlap_img_list)
class TmpImage(ut.NiceRepr):
pass
from skimage.feature import hog
from skimage import data, color, exposure
import wbia.plottool as pt
image2 = color.rgb2gray(data.astronaut()) # NOQA
fpath = './GOPR1120.JPG'
import vtool as vt
for fpath in [fpath]:
"""
http://scikit-image.org/docs/dev/auto_examples/plot_hog.html
"""
image = vt.imread(fpath, grayscale=True)
image = pt.color_funcs.to_base01(image)
fig = pt.figure(fnum=2)
fd, hog_image = hog(
image,
orientations=8,
pixels_per_cell=(16, 16),
cells_per_block=(1, 1),
visualise=True,
)
fig, (ax1, ax2) = pt.plt.subplots(1,
2,
figsize=(8, 4),
sharex=True,
sharey=True)
ax1.axis('off')
ax1.imshow(image, cmap=pt.plt.cm.gray)
ax1.set_title('Input image')
ax1.set_adjustable('box-forced')
# Rescale histogram for better display
hog_image_rescaled = exposure.rescale_intensity(hog_image,
in_range=(0, 0.02))
ax2.axis('off')
ax2.imshow(hog_image_rescaled, cmap=pt.plt.cm.gray)
ax2.set_title('Histogram of Oriented Gradients')
ax1.set_adjustable('box-forced')
pt.plt.show()
def clean_mitex():
# if mitex does not install correctly
install_dir = 'C:\Program Files (x86)\MiKTeX 2.9'
ut.delete(install_dir)
def install_mathtools():
"""
wget http://mirrors.ctan.org/install/macros/latex/contrib/mathtools.tds.zip
mkdir tmp2
#7z x -o"tmp2" mathtools.tds.zip
7z x -o"C:/Program Files (x86)/MiKTeX 2.9" mathtools.tds.zip
"""
pass
if __name__ == '__main__':
"""
python %USERPROFILE%/local/windows/init_mitex.py
"""
assert ut.WIN32
url = 'http://mirrors.ctan.org/systems/win32/miktex/setup/basic-miktex-2.9.5105.exe'
fpath = ut.grab_file_url(url)
ut.cmd(fpath)
def _train_setup(dark, voc_path, weight_path):
class_list = []
annotations_path = join(voc_path, "Annotations")
imagesets_path = join(voc_path, "ImageSets")
jpegimages_path = join(voc_path, "JPEGImages")
label_path = join(voc_path, "labels")
ut.delete(label_path)
ut.ensuredir(label_path)
def _convert_annotation(image_id):
import xml.etree.ElementTree as ET
def _convert(size, box):
dw = 1.0 / size[0]
dh = 1.0 / size[1]
x = (box[0] + box[1]) / 2.0
y = (box[2] + box[3]) / 2.0
w = box[1] - box[0]
h = box[3] - box[2]
x = x * dw
w = w * dw
y = y * dh
h = h * dh
return (x, y, w, h)
with open(join(label_path, "%s.txt" % (image_id,)), "w") as out_file:
with open(join(annotations_path, "%s.xml" % (image_id,)), "r") as in_file:
tree = ET.parse(in_file)
root = tree.getroot()
size = root.find("size")
w = int(size.find("width").text)
h = int(size.find("height").text)
for obj in root.iter("object"):
if int(obj.find("difficult").text) == 1:
continue
class_ = obj.find("name").text
if class_ not in class_list:
class_list.append(class_)
class_id = class_list.index(class_)
xmlbox = obj.find("bndbox")
b = tuple(
map(
float,
[
xmlbox.find("xmin").text,
xmlbox.find("xmax").text,
xmlbox.find("ymin").text,
xmlbox.find("ymax").text,
],
)
)
bb = _convert((w, h), b)
bb_str = " ".join([str(_) for _ in bb])
out_file.write("%s %s\n" % (class_id, bb_str))
num_images = 0
print("[pydarknet py train] Processing manifest...")
manifest_filename = join(voc_path, "manifest.txt")
with open(manifest_filename, "w") as manifest:
# for dataset_name in ['train', 'val', 'test']:
for dataset_name in ["train", "val"]:
dataset_filename = join(imagesets_path, "Main", "%s.txt" % dataset_name)
with open(dataset_filename, "r") as dataset:
image_id_list = dataset.read().strip().split()
for image_id in image_id_list:
print("[pydarknet py train] processing: %r" % (image_id,))
image_filepath = abspath(join(jpegimages_path, "%s.jpg" % image_id))
if exists(image_filepath):
manifest.write("%s\n" % (image_filepath,))
_convert_annotation(image_id)
num_images += 1
print("[pydarknet py train] Processing config and pretrained weights...")
# Load default config and pretrained weights
config_filepath = ut.grab_file_url(DEFAULT_CONFIG_TEMPLATE_URL, appname="pydarknet")
with open(config_filepath, "r") as config:
config_template_str = config.read()
config_filename = basename(config_filepath).replace(".template.", ".%d." % (len(class_list),))
config_filepath = join(weight_path, config_filename)
with open(config_filepath, "w") as config:
replace_list = [
("_^_OUTPUT_^_", SIDES * SIDES * (BOXES * 5 + len(class_list))),
("_^_CLASSES_^_", len(class_list)),
("_^_SIDES_^_", SIDES),
("_^_BOXES_^_", BOXES),
]
for needle, replacement in replace_list:
config_template_str = config_template_str.replace(needle, str(replacement))
config.write(config_template_str)
class_filepath = "%s.classes" % (config_filepath,)
with open(class_filepath, "w") as class_file:
for class_ in class_list:
class_file.write("%s\n" % (class_,))
weight_filepath = ut.grab_file_url(DEFAULT_PRETRAINED_URL, appname="pydarknet")
dark._load(config_filepath, weight_filepath)
print("class_list = %r" % (class_list,))
print("num_images = %r" % (num_images,))
return manifest_filename, num_images, config_filepath, class_filepath
def detect(gpath_list,
config_filepath,
weight_filepath,
class_filepath,
sensitivity,
verbose=VERBOSE_LN,
**kwargs):
"""Detect image filepaths with lightnet.
Args:
gpath_list (list of str): the list of image paths that need proposal candidates
Kwargs (optional): refer to the Lightnet documentation for configuration settings
Returns:
iter
"""
assert config_filepath is None, 'lightnet does not have a config file'
# Get correct weight if specified with shorthand
weight_url = None
if weight_filepath in WEIGHT_URL_DICT:
weight_url = WEIGHT_URL_DICT[weight_filepath]
weight_filepath = ut.grab_file_url(weight_url,
appname='ibeis',
check_hash=True)
if class_filepath in WEIGHT_URL_DICT:
if class_filepath is None and weight_url is not None:
weight_url_ = weight_url
else:
weight_url_ = WEIGHT_URL_DICT[weight_filepath]
class_url = _parse_classes_from_weights(weight_url_)
class_filepath = ut.grab_file_url(class_url,
appname='ibeis',
check_hash=True,
verbose=verbose)
assert exists(weight_filepath)
weight_filepath = ut.truepath(weight_filepath)
assert exists(class_filepath)
class_filepath = ut.truepath(class_filepath)
class_list = _parse_class_list(class_filepath)
network_size = (416, 416)
conf_thresh = sensitivity
nms_thresh = 1.0 # Turn off NMS
network = _create_network(weight_filepath, class_list, conf_thresh,
nms_thresh, network_size)
# Execute detector for each image
results_list_ = []
for gpath in tqdm(gpath_list):
image, output_list = _detect(network, gpath, network_size)
output_list = output_list[0]
result_list_ = []
for output in list(output_list):
xtl = int(np.around(float(output.x_top_left)))
ytl = int(np.around(float(output.y_top_left)))
xbr = int(np.around(float(output.x_top_left + output.width)))
ybr = int(np.around(float(output.y_top_left + output.height)))
class_ = output.class_label
conf = float(output.confidence)
result_dict = {
'xtl': xtl,
'ytl': ytl,
'width': xbr - xtl,
'height': ybr - ytl,
'class': class_,
'confidence': conf,
}
result_list_.append(result_dict)
results_list_.append(result_list_)
if len(results_list_) != len(gpath_list):
raise ValueError('Lightnet did not return valid data')
results_list = zip(gpath_list, results_list_)
return results_list
def test(
gpath_list,
classifier_weight_filepath=None,
return_dict=False,
multiclass=False,
**kwargs,
):
from wbia.detecttools.directory import Directory
# Get correct weight if specified with shorthand
archive_url = None
ensemble_index = None
if classifier_weight_filepath is not None and ':' in classifier_weight_filepath:
assert classifier_weight_filepath.count(':') == 1
classifier_weight_filepath, ensemble_index = classifier_weight_filepath.split(
':')
ensemble_index = int(ensemble_index)
if classifier_weight_filepath in ARCHIVE_URL_DICT:
archive_url = ARCHIVE_URL_DICT[classifier_weight_filepath]
archive_path = ut.grab_file_url(archive_url,
appname='wbia',
check_hash=True)
else:
logger.info('classifier_weight_filepath %r not recognized' %
(classifier_weight_filepath, ))
raise RuntimeError
assert os.path.exists(archive_path)
archive_path = ut.truepath(archive_path)
ensemble_path = archive_path.strip('.zip')
if not os.path.exists(ensemble_path):
ut.unarchive_file(archive_path, output_dir=ensemble_path)
assert os.path.exists(ensemble_path)
direct = Directory(ensemble_path,
include_file_extensions=['weights'],
recursive=True)
weights_path_list = direct.files()
weights_path_list = sorted(weights_path_list)
assert len(weights_path_list) > 0
kwargs.pop('classifier_algo', None)
logger.info('Using weights in the ensemble, index %r: %s ' %
(ensemble_index, ut.repr3(weights_path_list)))
result_list = test_ensemble(
gpath_list,
weights_path_list,
classifier_weight_filepath,
ensemble_index,
multiclass=multiclass,
**kwargs,
)
for result in result_list:
best_key = None
best_score = -1.0
for key, score in result.items():
if score > best_score:
best_key = key
best_score = score
assert best_score >= 0.0 and best_key is not None
if return_dict:
yield best_score, best_key, result
else:
yield best_score, best_key
def detect(
gpath_list,
config_filepath=None,
weight_filepath=None,
classes_filepath=None,
sensitivity=0.0,
verbose=VERBOSE_LN,
flip=False,
batch_size=192,
**kwargs,
):
"""Detect image filepaths with lightnet.
Args:
gpath_list (list of str): the list of image paths that need proposal candidates
Kwargs (optional): refer to the Lightnet documentation for configuration settings
Returns:
iter
"""
# Get correct weight if specified with shorthand
config_url = None
if config_filepath in CONFIG_URL_DICT:
config_url = CONFIG_URL_DICT[config_filepath]
config_filepath = ut.grab_file_url(config_url,
appname='lightnet',
check_hash=True)
# Get correct weights if specified with shorthand
if weight_filepath in CONFIG_URL_DICT:
if weight_filepath is None and config_url is not None:
config_url_ = config_url
else:
config_url_ = CONFIG_URL_DICT[weight_filepath]
weight_url = _parse_weights_from_cfg(config_url_)
weight_filepath = ut.grab_file_url(weight_url,
appname='lightnet',
check_hash=True)
assert exists(config_filepath)
config_filepath = ut.truepath(config_filepath)
assert exists(weight_filepath)
weight_filepath = ut.truepath(weight_filepath)
conf_thresh = sensitivity
nms_thresh = 1.0 # Turn off NMS
params = _create_network(config_filepath, weight_filepath, conf_thresh,
nms_thresh)
# Execute detector for each image
results_list_ = []
for gpath_batch_list in tqdm(list(ut.ichunks(gpath_list, batch_size))):
try:
result_list, img_sizes = _detect(params,
gpath_batch_list,
flip=flip)
except cv2.error:
result_list, img_sizes = [], []
for result, img_size in zip(result_list, img_sizes):
img_w, img_h = img_size
result_list_ = []
for output in list(result):
xtl = int(np.around(float(output.x_top_left)))
ytl = int(np.around(float(output.y_top_left)))
xbr = int(np.around(float(output.x_top_left + output.width)))
ybr = int(np.around(float(output.y_top_left + output.height)))
width = xbr - xtl
height = ybr - ytl
class_ = output.class_label
conf = float(output.confidence)
if flip:
xtl = img_w - xbr
result_dict = {
'xtl': xtl,
'ytl': ytl,
'width': width,
'height': height,
'class': class_,
'confidence': conf,
}
result_list_.append(result_dict)
results_list_.append(result_list_)
if len(results_list_) != len(gpath_list):
raise ValueError('Lightnet did not return valid data')
results_list = zip(gpath_list, results_list_)
return results_list
ytl = ymin
w = xmax - xmin
h = ymax - ymin
return (
xtl,
ytl,
w,
h,
)
test_uuid_list = []
for model_tag in TEST_URLS:
json_url = TEST_URLS[model_tag]
json_filepath = ut.grab_file_url(json_url,
appname='wbia_2d_orientation',
check_hash=False)
with open(json_filepath, 'r') as file:
json_data = yaml.load(file, Loader=yaml.FullLoader)
annotation_data = json_data['annotations']
test_uuid_list_ = ut.take_column(annotation_data, 'uuid')
print(len(set(test_uuid_list_)))
test_uuid_list += test_uuid_list_
assert len(test_uuid_list) == len(set(test_uuid_list))
test_uuid_list = [uuid.UUID(test_uuid) for test_uuid in test_uuid_list]
test_uuid_set = set(test_uuid_list)
################################################################################
desired_species_list = [
def ingest_serengeti_mamal_cameratrap(species):
"""
Downloads data from Serengeti dryad server
References:
http://datadryad.org/resource/doi:10.5061/dryad.5pt92
Swanson AB, Kosmala M, Lintott CJ, Simpson RJ, Smith A, Packer C (2015)
Snapshot Serengeti, high-frequency annotated camera trap images of 40
mammalian species in an African savanna. Scientific Data 2: 150026.
http://dx.doi.org/10.1038/sdata.2015.26
Swanson AB, Kosmala M, Lintott CJ, Simpson RJ, Smith A, Packer C (2015)
Data from: Snapshot Serengeti, high-frequency annotated camera trap
images of 40 mammalian species in an African savanna. Dryad Digital
Repository. http://dx.doi.org/10.5061/dryad.5pt92
Args:
species (?):
CommandLine:
python -m ibeis.dbio.ingest_database --test-ingest_serengeti_mamal_cameratrap --species zebra_plains
python -m ibeis.dbio.ingest_database --test-ingest_serengeti_mamal_cameratrap --species cheetah
Example:
>>> # SCRIPT
>>> from ibeis.dbio.ingest_database import * # NOQA
>>> import ibeis
>>> species = ut.get_argval('--species', type_=str, default=ibeis.const.TEST_SPECIES.ZEB_PLAIN)
>>> # species = ut.get_argval('--species', type_=str, default='cheetah')
>>> result = ingest_serengeti_mamal_cameratrap(species)
>>> print(result)
"""
'https://snapshotserengeti.s3.msi.umn.edu/'
import ibeis
if species is None:
code = 'ALL'
elif species == 'zebra_plains':
code = 'PZ'
elif species == 'cheetah':
code = 'CHTH'
else:
raise NotImplementedError()
if species == 'zebra_plains':
serengeti_sepcies = 'zebra'
else:
serengeti_sepcies = species
print('species = %r' % (species,))
print('serengeti_sepcies = %r' % (serengeti_sepcies,))
dbname = code + '_Serengeti'
print('dbname = %r' % (dbname,))
dbdir = ut.ensuredir(join(ibeis.sysres.get_workdir(), dbname))
print('dbdir = %r' % (dbdir,))
image_dir = ut.ensuredir(join(dbdir, 'images'))
base_url = 'http://datadryad.org/bitstream/handle/10255'
all_images_url = base_url + '/dryad.86392/all_images.csv'
consensus_metadata_url = base_url + '/dryad.86348/consensus_data.csv'
search_effort_url = base_url + '/dryad.86347/search_effort.csv'
gold_standard_url = base_url + '/dryad.76010/gold_standard_data.csv'
all_images_fpath = ut.grab_file_url(all_images_url, download_dir=dbdir)
consensus_metadata_fpath = ut.grab_file_url(consensus_metadata_url, download_dir=dbdir)
search_effort_fpath = ut.grab_file_url(search_effort_url, download_dir=dbdir)
gold_standard_fpath = ut.grab_file_url(gold_standard_url, download_dir=dbdir)
print('all_images_fpath = %r' % (all_images_fpath,))
print('consensus_metadata_fpath = %r' % (consensus_metadata_fpath,))
print('search_effort_fpath = %r' % (search_effort_fpath,))
print('gold_standard_fpath = %r' % (gold_standard_fpath,))
def read_csv(csv_fpath):
import utool as ut
csv_text = ut.read_from(csv_fpath)
csv_lines = csv_text.split('\n')
print(ut.list_str(csv_lines[0:2]))
csv_data = [[field.strip('"').strip('\r') for field in line.split(',')]
for line in csv_lines if len(line) > 0]
csv_header = csv_data[0]
csv_data = csv_data[1:]
return csv_data, csv_header
def download_image_urls(image_url_info_list):
# Find ones that we already have
print('Requested %d downloaded images' % (len(image_url_info_list)))
full_gpath_list = [join(image_dir, basename(gpath)) for gpath in image_url_info_list]
exists_list = [ut.checkpath(gpath) for gpath in full_gpath_list]
image_url_info_list_ = ut.compress(image_url_info_list, ut.not_list(exists_list))
print('Already have %d/%d downloaded images' % (
len(image_url_info_list) - len(image_url_info_list_), len(image_url_info_list)))
print('Need to download %d images' % (len(image_url_info_list_)))
#import sys
#sys.exit(0)
# Download the rest
imgurl_prefix = 'https://snapshotserengeti.s3.msi.umn.edu/'
image_url_list = [imgurl_prefix + suffix for suffix in image_url_info_list_]
for img_url in ut.ProgressIter(image_url_list, lbl='Downloading image'):
ut.grab_file_url(img_url, download_dir=image_dir)
return full_gpath_list
# Data contains information about which events have which animals
if False:
species_class_csv_data, species_class_header = read_csv(gold_standard_fpath)
species_class_eventid_list = ut.get_list_column(species_class_csv_data, 0)
#gold_num_species_annots_list = ut.get_list_column(gold_standard_csv_data, 2)
species_class_species_list = ut.get_list_column(species_class_csv_data, 2)
#gold_count_list = ut.get_list_column(gold_standard_csv_data, 3)
else:
species_class_csv_data, species_class_header = read_csv(consensus_metadata_fpath)
species_class_eventid_list = ut.get_list_column(species_class_csv_data, 0)
species_class_species_list = ut.get_list_column(species_class_csv_data, 7)
# Find the zebra events
serengeti_sepcies_set = sorted(list(set(species_class_species_list)))
print('serengeti_sepcies_hist = %s' %
ut.dict_str(ut.dict_hist(species_class_species_list), key_order_metric='val'))
#print('serengeti_sepcies_set = %s' % (ut.list_str(serengeti_sepcies_set),))
assert serengeti_sepcies in serengeti_sepcies_set, 'not a known seregeti species'
species_class_chosen_idx_list = ut.list_where(
[serengeti_sepcies == species_ for species_ in species_class_species_list])
chosen_eventid_list = ut.take(species_class_eventid_list, species_class_chosen_idx_list)
print('Number of chosen species:')
print(' * len(species_class_chosen_idx_list) = %r' % (len(species_class_chosen_idx_list),))
print(' * len(chosen_eventid_list) = %r' % (len(chosen_eventid_list),))
# Read info about which events have which images
images_csv_data, image_csv_header = read_csv(all_images_fpath)
capture_event_id_list = ut.get_list_column(images_csv_data, 0)
image_url_info_list = ut.get_list_column(images_csv_data, 1)
# Group photos by eventid
eventid_to_photos = ut.group_items(image_url_info_list, capture_event_id_list)
# Filter to only chosens
unflat_chosen_url_infos = ut.dict_take(eventid_to_photos, chosen_eventid_list)
chosen_url_infos = ut.flatten(unflat_chosen_url_infos)
image_url_info_list = chosen_url_infos
chosen_path_list = download_image_urls(chosen_url_infos)
ibs = ibeis.opendb(dbdir=dbdir, allow_newdir=True)
gid_list_ = ibs.add_images(chosen_path_list, auto_localize=False) # NOQA
# Attempt to automatically detect the annotations
#aids_list = ibs.detect_random_forest(gid_list_, species)
#aids_list
#if False:
# # remove non-zebra photos
# from os.path import basename
# base_gname_list = list(map(basename, zebra_url_infos))
# all_gname_list = ut.list_images(image_dir)
# nonzebra_gname_list = ut.setdiff_ordered(all_gname_list, base_gname_list)
# nonzebra_gpath_list = ut.fnames_to_fpaths(nonzebra_gname_list, image_dir)
# ut.remove_fpaths(nonzebra_gpath_list)
return ibs
def post(self):
global NETWORK_MODEL_TAG
global NETWORK
global NETWORK_VALUES
response = {'success': False}
# ut.embed()
try:
with ut.Timer('Pre'):
parser = reqparse.RequestParser()
parser.add_argument('image', type=str)
parser.add_argument('config', type=dict)
args = parser.parse_args()
image_base64_str = args['image']
image = get_image_from_base64_str(image_base64_str)
config = args['config']
model_tag = config.get('model_tag', None)
num_returns = config.get('topk', 100)
model_url = model_url_dict.get(model_tag, None)
assert model_url is not None, 'Model tag %r is not recognized' % (
model_tag, )
if model_tag != NETWORK_MODEL_TAG:
with ut.Timer('Loading network'):
print('Loading network from weights %r' % (model_tag, ))
values_url = model_url.replace('.pth', '.values.pth')
# Download files
model_filepath = ut.grab_file_url(model_url,
appname='kaggle7',
check_hash=True)
values_filepath = ut.grab_file_url(values_url,
appname='kaggle7',
check_hash=True)
model_values = torch.load(values_filepath)
classes = model_values['classes']
num_classes = len(classes)
model_weights = torch.load(model_filepath,
map_location=get_device())
network_model, mutliple = make_new_network(
num_classes, RING_HEADS, GEM_CONST, pretrained=False)
if mutliple:
pass
if torch.cuda.is_available():
network_model = network_model.cuda()
# model_weights = model_weights['model']
network_model.load_state_dict(model_weights)
network_model.eval()
NETWORK_MODEL_TAG = model_tag
NETWORK = network_model
NETWORK_VALUES = model_values
print('Using network %r' % (NETWORK_MODEL_TAG, ))
with ut.Timer('Loading input tensor'):
input_image = image.convert(CMODE).convert('LA').convert(CMODE)
input_image = TFRM_RESIZE(input_image)
input_image = pil2tensor(input_image, np.float32)
input_image = input_image.div_(255)
input_image = TFRM_WHITEN(input_image)
size = input_image.size()
input_tensor = input_image.view(-1, size[0], size[1], size[2])
input_tensor = input_tensor.to(get_device())
# Run inference
with ut.Timer('Inference'):
print('Running inference on input tensor %r' %
(input_tensor.size(), ))
output = NETWORK(input_tensor)
print('...done')
preds_list, feats_list = output
with ut.Timer('Post1'):
print('Performing post-processing')
prediction_raw = preds_list[-1][0]
features_raw = TFRM_L2NORM(torch.cat(feats_list, dim=1))[0]
with ut.Timer('Post2'):
print('...classifier')
# Post Process classification
classifier_temp = NETWORK_VALUES['thresholds'][
'classifier_softmax_temp']
classifier_prediction = torch.softmax(prediction_raw /
classifier_temp,
dim=0)
with ut.Timer('Post3'):
# Post process features
print('...features')
train_feats = NETWORK_VALUES['train_feats']
train_gt = NETWORK_VALUES['train_gt']
size = features_raw.size()
features = features_raw.view(-1, size[0])
distance_matrix_imgs = batched_dmv(features, train_feats)
distance_matrix_classes = dm2cm(distance_matrix_imgs, train_gt)
features_sim = (2.0 - distance_matrix_classes) * 0.5
features_sim = features_sim[0]
features_temp = NETWORK_VALUES['thresholds'][
'feature_softmax_temp']
features_prediction = torch.softmax(features_sim /
features_temp,
dim=0)
with ut.Timer('Post4'):
print('...mixing')
p = NETWORK_VALUES['thresholds']['mixing_value']
classifier_prediction = classifier_prediction.to('cpu')
final_prediction = (p * classifier_prediction +
(1.0 - p) * features_prediction)
with ut.Timer('Collection'):
print('Collecting prediction')
top_k_score_list, top_k_index_list = final_prediction.topk(
num_returns, 0)
top_k_score_list = top_k_score_list.detach().tolist()
classes = NETWORK_VALUES['classes']
top_k_class_list = ut.take(classes, top_k_index_list)
response['scores'] = {}
for top_k_class, top_k_score in zip(top_k_class_list,
top_k_score_list):
response['scores'][top_k_class] = top_k_score
response['success'] = True
print('...done')
except Exception as ex:
message = str(ex)
response['message'] = message
print('!!!ERROR!!!')
print(response)
# if torch.cuda.is_available():
# torch.cuda.empty_cache()
return response
def classifyShark(ibs, gid_list):
suffix = 'lenet'
batch_size = 32
model_name = 'injur-shark-' + suffix
model = classify_shark.WhaleSharkInjuryModel(name=model_name,
output_dims=2,
data_shape=(224, 224, 3),
batch_size=batch_size)
model.init_arch()
filep = ut.grab_file_url(modelStateLocation)
model.load_model_state(fpath=filep)
model.rrr()
config = {
'algo': 'yolo',
'sensitivity': 0.2,
'config_filepath': 'default',
}
depc = ibs.depc_image
images = ibs.images(gid_list)
images = images.compress([ext_ not in ['.gif'] for ext_ in images.exts])
gid_list = images.gids
# uuid_gid_list = [str(item) for item in images.uuids]
results_list = depc.get_property('localizations',
gid_list,
None,
config=config) # NOQA
results_list2 = []
multi_gids = []
failed_gids = []
for gid, res in zip(gid_list, results_list):
score, bbox_list, theta_list, conf_list, class_list = res
if len(bbox_list) == 0:
failed_gids.append(gid)
elif len(bbox_list) == 1:
results_list2.append((gid, bbox_list, theta_list))
elif len(bbox_list) > 1:
# Take only a single annotation per bounding box.
multi_gids.append(gid)
idx = conf_list.argmax()
res2 = (gid, bbox_list[idx:idx + 1], theta_list[idx:idx + 1])
results_list2.append(res2)
# Reorder empty_info to be aligned with results
localized_imgs = ibs.images(ut.take_column(results_list2, 0))
# Override old bboxes
bboxes = np.array(ut.take_column(results_list2, 1))[:, 0, :]
thetas = np.array(ut.take_column(results_list2, 2))[:, 0]
species = ['whale_shark'] * len(localized_imgs)
aid_list = ibs.add_annots(localized_imgs.gids,
bbox_list=bboxes,
theta_list=thetas,
species_list=species)
config = {'dim_size': (224, 224), 'resize_dim': 'wh'}
chip_gen = ibs.depc_annot.get('chips',
aid_list,
'img',
eager=False,
config=config)
data_shape = config['dim_size'] + (3, )
iter_ = iter(ut.ProgIter(chip_gen, nTotal=len(aid_list), lbl='load chip'))
shape = (len(aid_list), ) + data_shape
data = vt.fromiter_nd(iter_, shape=shape, dtype=np.uint8) # NOQA
results = model._predict(data)
predictions = results['predictions']
classes = np.array(['healthy', 'injured'])
prediction_class = classes[np.array(predictions)]
return {
'predictions': prediction_class.tolist(),
'confidences': results['confidences'].tolist(),
}