def my_main(tag, cfg_file, set_cfgs, imdb_name, _config):
# Already set everything here, so the path can be determined correctly
if cfg_file:
cfg_from_file(cfg_file)
if set_cfgs:
cfg_from_list(set_cfgs)
print('Called with args:')
print(_config)
# if not already present save the configuration into a file in the output folder
outdir = osp.abspath(
osp.join(cfg.ROOT_DIR, 'output', 'frcnn', cfg.EXP_DIR, imdb_name, tag))
sacred_config = osp.join(outdir, 'sacred_config.yaml')
if not osp.isfile(sacred_config):
if not os.path.exists(outdir):
os.makedirs(outdir)
with open(sacred_config, 'w') as outfile:
yaml.dump({'frcnn': _config}, outfile, default_flow_style=False)
frcnn_trainval()
def my_main(imdbtest_name, clip_bbox, output_name, nms_thresh, frcnn,
write_images, _config):
# Clip bboxes after bbox reg to image boundary
cfg_from_list(
['TEST.BBOX_CLIP',
str(clip_bbox), 'TEST.NMS',
str(nms_thresh)])
#cfg_from_list(["APPLY_CLAHE", "True"])
# Already set everything here, so the path can be determined correctly
if frcnn['cfg_file']:
cfg_from_file(frcnn['cfg_file'])
if frcnn['set_cfgs']:
cfg_from_list(frcnn['set_cfgs'])
model_dir = osp.abspath(
osp.join(cfg.ROOT_DIR, 'output', 'frcnn', cfg.EXP_DIR,
frcnn['imdb_name'], frcnn['tag']))
model = osp.join(
model_dir, cfg.TRAIN.SNAPSHOT_PREFIX +
'_iter_{:d}'.format(frcnn['max_iters']) + '.pth')
# model = osp.join(model_dir, cfg.TRAIN.SNAPSHOT_PREFIX + '_iter_35000' + '.pth')
if output_name:
output_dir = osp.join(model_dir, output_name)
else:
output_dir = osp.join(model_dir, imdbtest_name)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
print('Called with args:')
print(_config)
frcnn_test(model=model,
output_dir=output_dir,
network=frcnn['network'],
write_images=write_images)
from __future__ import absolute_import, division, print_function
import os
import os.path as osp
# Change paths in config file
from frcnn.model import config
from frcnn.model.config import cfg, cfg_from_file, cfg_from_list
from frcnn_test import frcnn_test
from sacred import Experiment
this_dir = osp.dirname(__file__)
root_path = osp.join(this_dir, '..', '..')
data_path = osp.join(root_path, 'data')
config.cfg_from_list(['ROOT_DIR', root_path, 'DATA_DIR', data_path])
ex = Experiment()
frcnn_test = ex.capture(frcnn_test)
@ex.config
def default():
score_thresh = 0.05
nms_thresh = 0.3
clip_bbox = False
max_per_image = 100
output_name = None
write_images = False
# Added so that sacred doesn't throw a key error
def my_main(tracktor, siamese, _config):
# set all seeds
torch.manual_seed(tracktor['seed'])
torch.cuda.manual_seed(tracktor['seed'])
np.random.seed(tracktor['seed'])
torch.backends.cudnn.deterministic = True
output_dir = osp.join(get_output_dir(tracktor['module_name']),
tracktor['name'])
sacred_config = osp.join(output_dir, 'sacred_config.yaml')
if not osp.exists(output_dir):
os.makedirs(output_dir)
with open(sacred_config, 'w') as outfile:
yaml.dump(_config, outfile, default_flow_style=False)
##########################
# Initialize the modules #
##########################
# object detection
print("[*] Building object detector")
if tracktor['network'].startswith('frcnn'):
# FRCNN
from tracktor.frcnn import FRCNN
from frcnn.model import config
if _config['frcnn']['cfg_file']:
config.cfg_from_file(_config['frcnn']['cfg_file'])
if _config['frcnn']['set_cfgs']:
config.cfg_from_list(_config['frcnn']['set_cfgs'])
obj_detect = FRCNN(num_layers=101)
obj_detect.create_architecture(2,
tag='default',
anchor_scales=config.cfg.ANCHOR_SCALES,
anchor_ratios=config.cfg.ANCHOR_RATIOS)
obj_detect.load_state_dict(torch.load(tracktor['obj_detect_weights']))
elif tracktor['network'].startswith('fpn'):
# FPN
from tracktor.fpn import FPN
from fpn.model.utils import config
config.cfg.TRAIN.USE_FLIPPED = False
config.cfg.CUDA = True
config.cfg.TRAIN.USE_FLIPPED = False
checkpoint = torch.load(tracktor['obj_detect_weights'])
if 'pooling_mode' in checkpoint.keys():
config.cfg.POOLING_MODE = checkpoint['pooling_mode']
set_cfgs = [
'ANCHOR_SCALES', '[4, 8, 16, 32]', 'ANCHOR_RATIOS', '[0.5,1,2]'
]
config.cfg_from_file(_config['tracktor']['obj_detect_config'])
config.cfg_from_list(set_cfgs)
obj_detect = FPN(('__background__', 'pedestrian'),
101,
pretrained=False)
obj_detect.create_architecture()
obj_detect.load_state_dict(checkpoint['model'])
else:
raise NotImplementedError(
f"Object detector type not known: {tracktor['network']}")
pprint.pprint(config.cfg)
obj_detect.eval()
obj_detect.cuda()
# reid
reid_network = resnet50(pretrained=False, **siamese['cnn'])
reid_network.load_state_dict(torch.load(tracktor['reid_network_weights']))
reid_network.eval()
reid_network.cuda()
# tracktor
if 'oracle' in tracktor:
tracker = OracleTracker(obj_detect, reid_network, tracktor['tracker'],
tracktor['oracle'])
else:
tracker = Tracker(obj_detect, reid_network, tracktor['tracker'])
print("[*] Beginning evaluation...")
time_total = 0
for sequence in Datasets(tracktor['dataset']):
tracker.reset()
now = time.time()
print("[*] Evaluating: {}".format(sequence))
data_loader = DataLoader(sequence, batch_size=1, shuffle=False)
for i, frame in enumerate(data_loader):
# frame_split = [0.0, 1.0]
if i >= len(sequence) * tracktor['frame_split'][0] and i <= len(
sequence) * tracktor['frame_split'][1]:
tracker.step(frame)
results = tracker.get_results()
time_total += time.time() - now
print("[*] Tracks found: {}".format(len(results)))
print("[*] Time needed for {} evaluation: {:.3f} s".format(
sequence,
time.time() - now))
if tracktor['interpolate']:
results = interpolate(results)
sequence.write_results(results, osp.join(output_dir))
if tracktor['write_images']:
plot_sequence(
results, sequence,
osp.join(output_dir, tracktor['dataset'], str(sequence)))
print("[*] Evaluation for all sets (without image generation): {:.3f} s".
format(time_total))
def my_main(tracktor, siamese, _config):
# set all seeds
torch.manual_seed(tracktor['seed'])
torch.cuda.manual_seed(tracktor['seed'])
np.random.seed(tracktor['seed'])
torch.backends.cudnn.deterministic = True
output_dir = osp.join(get_output_dir(tracktor['module_name']),
tracktor['name'])
sacred_config = osp.join(output_dir, 'sacred_config.yaml')
if not osp.exists(output_dir):
os.makedirs(output_dir)
with open(sacred_config, 'w') as outfile:
yaml.dump(_config, outfile, default_flow_style=False)
##########################
# Initialize the modules #
##########################
# object detection
print("[*] Building object detector")
print("tracktor['network'] is: ", tracktor['network'])
if tracktor['network'].startswith('frcnn'):
# FRCNN
from tracktor.frcnn import FRCNN
from frcnn.model import config
if _config['frcnn']['cfg_file']:
config.cfg_from_file(_config['frcnn']['cfg_file'])
if _config['frcnn']['set_cfgs']:
config.cfg_from_list(_config['frcnn']['set_cfgs'])
obj_detect = FRCNN(num_layers=101)
obj_detect.create_architecture(2,
tag='default',
anchor_scales=config.cfg.ANCHOR_SCALES,
anchor_ratios=config.cfg.ANCHOR_RATIOS)
state_dict_person = torch.load(tracktor['obj_detect_weights_person'])
obj_detect.load_state_dict(state_dict_person)
# loading head-detection model
obj_detect_head = FRCNN(num_layers=101)
obj_detect_head.create_architecture(
2,
tag='default',
anchor_scales=config.cfg.ANCHOR_SCALES,
anchor_ratios=config.cfg.ANCHOR_RATIOS)
state_dict_head = torch.load(tracktor['obj_detect_weights_head'])
state_dict_head = my_transform(state_dict_head)
obj_detect_head.load_state_dict(state_dict_head)
elif tracktor['network'].startswith('mask-rcnn'):
# MASK-RCNN
pass
elif tracktor['network'].startswith('fpn'):
# FPN
from tracktor.fpn import FPN
from fpn.model.utils import config
config.cfg.TRAIN.USE_FLIPPED = False
config.cfg.CUDA = True
config.cfg.TRAIN.USE_FLIPPED = False
checkpoint = torch.load(tracktor['obj_detect_weights'])
if 'pooling_mode' in checkpoint.keys():
config.cfg.POOLING_MODE = checkpoint['pooling_mode']
set_cfgs = [
'ANCHOR_SCALES', '[4, 8, 16, 32]', 'ANCHOR_RATIOS', '[0.5,1,2]'
]
config.cfg_from_file(_config['tracktor']['obj_detect_config'])
config.cfg_from_list(set_cfgs)
obj_detect = FPN(('__background__', 'pedestrian'),
101,
pretrained=False)
obj_detect.create_architecture()
obj_detect.load_state_dict(checkpoint['model'])
else:
raise NotImplementedError(
f"Object detector type not known: {tracktor['network']}")
pprint.pprint(config.cfg)
obj_detect.eval()
obj_detect.cuda()
obj_detect_head.eval()
obj_detect_head.cuda()
# reid
reid_network = resnet50(pretrained=False, **siamese['cnn'])
reid_network.load_state_dict(torch.load(tracktor['reid_network_weights']))
reid_network.eval()
reid_network.cuda()
# tracktor
if 'oracle' in tracktor:
tracker = OracleTracker(obj_detect, reid_network, tracktor['tracker'],
tracktor['oracle'])
else:
print(tracktor['tracker'])
tracker = Tracker(obj_detect, reid_network, tracktor['tracker'])
print("[*] Beginning evaluation...")
time_total = 0
tracker.reset()
now = time.time()
cv2.namedWindow("test", cv2.WINDOW_NORMAL)
cv2.resizeWindow("test", 800, 600)
seq_name = 'MOT-2'
video_file = osp.join(cfg.ROOT_DIR, 'video/' + seq_name + '.mp4')
print("[*] Evaluating: {}".format(video_file))
# ===============================================
# transform each video frame to main frame format
# ===============================================
transforms = Compose(
[ToTensor(),
Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
vdo = cv2.VideoCapture()
vdo.open(video_file)
im_width = int(vdo.get(cv2.CAP_PROP_FRAME_WIDTH))
im_height = int(vdo.get(cv2.CAP_PROP_FRAME_HEIGHT))
area = 0, 0, im_width, im_height
print("===video frame's area:", area)
# video = cv2.VideoCapture(video_file)
# if not video.isOpened():
# print("error opening video stream or file!")
# while (video.isOpened()):
while vdo.grab():
_, frame = vdo.retrieve()
# success, frame = video.read()
# if not success:
# break
# print(frame) # (540, 960, 3)
blobs, im_scales = test._get_blobs(frame)
data = blobs['data']
# print(data.shape) # (1, 562, 1000, 3)
# print(im_scales) # [1.04166667]
sample = {}
sample['image'] = cv2.resize(frame, (0, 0),
fx=im_scales,
fy=im_scales,
interpolation=cv2.INTER_NEAREST)
sample['im_path'] = video_file
sample['data'] = torch.from_numpy(data).unsqueeze(0)
im_info = np.array([data.shape[1], data.shape[2], im_scales[0]],
dtype=np.float32)
sample['im_info'] = torch.from_numpy(im_info).unsqueeze(0)
# convert to siamese input
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame = Image.fromarray(frame)
frame = transforms(frame)
# print(frame.shape) # torch.Size([3, 540, 960])
sample['app_data'] = frame.unsqueeze(0).unsqueeze(0)
# print(sample['app_data'].size()) # torch.Size([1, 1, 3, 540, 960])
# additional info
# sample['gt'] = {}
# sample['vis'] = {}
# sample['dets'] = []
# print('frame begin')
# print(sample)
# print('frame end')
tracker.step(sample)
tracker.show_tracks(area)
video.release()
print('the current video' + video_file + ' is done')
results = tracker.get_results()
time_total += time.time() - now
print("[*] Tracks found: {}".format(len(results)))
print("[*] Time needed for {} evaluation: {:.3f} s".format(
seq_name,
time.time() - now))
# print('this is : ' + tracktor['dataset'])
# for sequence in Datasets(tracktor['dataset']):
# #for sequence in Datasets('MOT-02'):
#
# print('sequence---------', type(sequence), len(sequence))
#
# tracker.reset()
# now = time.time()
#
# print("[*] Evaluating: {}".format(sequence))
#
# data_loader = DataLoader(sequence, batch_size=1, shuffle=False)
# for i, frame in enumerate(data_loader):
#
# print('frame begin')
# print(frame)
# print('frame end')
#
# if i >= len(sequence) * tracktor['frame_split'][0] and i <= len(sequence) * tracktor['frame_split'][1]:
#
# tracker.step(frame)
# results = tracker.get_results()
#
#
# time_total += time.time() - now
#
# print("[*] Tracks found: {}".format(len(results)))
# print("[*] Time needed for {} evaluation: {:.3f} s".format(sequence, time.time() - now))
#
# if tracktor['interpolate']:
# results = interpolate(results)
#
# plot_tracks(sequence, results)
# sequence.write_results(results, osp.join(output_dir))
#
# if tracktor['write_images']:
# plot_sequence(results, sequence, osp.join(output_dir, tracktor['dataset'], str(sequence)))
print("[*] Evaluation for all sets (without image generation): {:.3f} s".
format(time_total))
def frcnn_trainval(imdb_name, imdbval_name, max_iters, pretrained_model,
pretrained_full_model, cfg_file, set_cfgs, network, tag):
"""
args = {'imdb_name':imdb_name,
'imdbval_name':imdbval_name,
'max_iters':max_iters,
'net':network,
'cfg_file':cfg_file,
'set_cfgs':set_cfgs,
'weights':weights,
'tag':tag}
"""
if cfg_file:
cfg_from_file(cfg_file)
if set_cfgs:
cfg_from_list(set_cfgs)
print('Using config:')
pprint.pprint(cfg)
np.random.seed(cfg.RNG_SEED)
# train set
imdb, roidb = combined_roidb(imdb_name)
print('{:d} roidb entries'.format(len(roidb)))
# output directory where the models are saved
output_dir = get_output_dir(imdb, tag)
print('Output will be saved to `{:s}`'.format(output_dir))
# tensorboard directory where the summaries are saved during training
tb_dir = get_output_tb_dir(imdb, tag)
print('TensorFlow summaries will be saved to `{:s}`'.format(tb_dir))
# also add the validation set, but with no flipping images
orgflip = cfg.TRAIN.USE_FLIPPED
cfg.TRAIN.USE_FLIPPED = False
#-, valroidb = combined_roidb(args['imdbval_name'])
valimdb, valroidb = combined_roidb(imdbval_name)
print('{:d} validation roidb entries'.format(len(valroidb)))
cfg.TRAIN.USE_FLIPPED = orgflip
# load network
if network == 'vgg16':
net = vgg16()
elif network == 'res50':
net = resnetv1(num_layers=50)
elif network == 'res101':
net = resnetv1(num_layers=101)
elif network == 'res152':
net = resnetv1(num_layers=152)
elif network == 'mobile':
net = mobilenetv1()
else:
raise NotImplementedError
train_net(net,
imdb,
roidb,
valroidb,
output_dir,
tb_dir,
pretrained_model=pretrained_model,
pretrained_full_model=pretrained_full_model,
max_iters=max_iters)
def my_main(tracktor, siamese, _config):
# set all seeds
torch.manual_seed(tracktor['seed'])
torch.cuda.manual_seed(tracktor['seed'])
np.random.seed(tracktor['seed'])
torch.backends.cudnn.deterministic = True
output_dir = osp.join(get_output_dir(tracktor['module_name']),
tracktor['name'])
sacred_config = osp.join(output_dir, 'sacred_config.yaml')
if not osp.exists(output_dir):
os.makedirs(output_dir)
with open(sacred_config, 'w') as outfile:
yaml.dump(_config, outfile, default_flow_style=False)
##########################
# Initialize the modules #
##########################
# object detection
print("[*] Building object detector")
if tracktor['network'].startswith('frcnn'):
# FRCNN
from tracktor.frcnn import FRCNN
from frcnn.model import config
if _config['frcnn']['cfg_file']:
config.cfg_from_file(_config['frcnn']['cfg_file'])
if _config['frcnn']['set_cfgs']:
config.cfg_from_list(_config['frcnn']['set_cfgs'])
obj_detect = FRCNN(num_layers=101)
obj_detect.create_architecture(2,
tag='default',
anchor_scales=config.cfg.ANCHOR_SCALES,
anchor_ratios=config.cfg.ANCHOR_RATIOS)
obj_detect.load_state_dict(torch.load(tracktor['obj_detect_weights']))
elif tracktor['network'].startswith('fpn'):
# FPN
from tracktor.fpn import FPN
from fpn.model.utils import config
config.cfg.TRAIN.USE_FLIPPED = False
config.cfg.CUDA = True
config.cfg.TRAIN.USE_FLIPPED = False
checkpoint = torch.load(tracktor['obj_detect_weights'])
if 'pooling_mode' in checkpoint.keys():
config.cfg.POOLING_MODE = checkpoint['pooling_mode']
set_cfgs = [
'ANCHOR_SCALES', '[4, 8, 16, 32]', 'ANCHOR_RATIOS', '[0.5,1,2]'
]
config.cfg_from_file(_config['tracktor']['obj_detect_config'])
config.cfg_from_list(set_cfgs)
obj_detect = FPN(('__background__', 'pedestrian'),
101,
pretrained=False)
obj_detect.create_architecture()
obj_detect.load_state_dict(checkpoint['model'])
else:
raise NotImplementedError(
f"Object detector type not known: {tracktor['network']}")
obj_detect.eval()
obj_detect.cuda()
print("[*] Beginning operation...")
layers = ['p2', 'p3', 'p4', 'p5']
f_hdf5 = h5py.File(
'/usr/stud/beckera/tracking_wo_bnw/data/motion/im_features.hdf5', 'w')
i_hdf5 = h5py.File(
'/usr/stud/beckera/tracking_wo_bnw/data/motion/images.hdf5', 'w')
for sequence in Datasets(tracktor['dataset']):
print("[*] Storing sequence: {}".format(sequence))
f_group = f_hdf5.create_group(sequence._seq_name)
i_group = i_hdf5.create_group(sequence._seq_name)
data_loader = DataLoader(sequence, batch_size=1, shuffle=False)
for i, frame in enumerate(data_loader):
if i == 0:
i_group.create_dataset('data',
shape=(len(data_loader),
*frame['data'][0].shape[1:]),
dtype='float16')
i_group.create_dataset('app_data',
shape=(len(data_loader),
*frame['app_data'][0].shape[1:]),
dtype='float16')
i_group.create_dataset('im_info',
shape=(len(data_loader), 3),
dtype='float16')
i_group['data'][i] = frame['data'][0].cpu().numpy()
i_group['app_data'][i] = frame['app_data'][0].cpu().numpy()
i_group['im_info'][i] = frame['im_info'].cpu().numpy()
image = Variable(frame['data'][0].permute(0, 3, 1, 2).cuda(),
volatile=True)
features = obj_detect.get_features(image)
for j, layer in enumerate(layers):
if i == 0:
f_group.create_dataset(layer,
shape=(len(data_loader),
*features[j].shape[1:]),
dtype='float16')
f_group[layer][i] = features[j].data.cpu().numpy().astype(
'float16')
f_hdf5.close()
i_hdf5.close()
imdb = datasets.imdb.imdb(imdb_names, tmp.classes)
else:
imdb = get_imdb(imdb_names)
return imdb, roidb
if __name__ == '__main__':
args = parse_args()
print('Called with args:')
print(args)
if args.cfg_file is not None:
cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs)
print('Using config:')
pprint.pprint(cfg)
np.random.seed(cfg.RNG_SEED)
# train set
imdb, roidb = combined_roidb(args.imdb_name)
print('{:d} roidb entries'.format(len(roidb)))
# output directory where the models are saved
output_dir = get_output_dir(imdb, args.tag)
print('Output will be saved to `{:s}`'.format(output_dir))
# tensorboard directory where the summaries are saved during training
def main(tracktor, siamese, _config):
# set all seeds
torch.manual_seed(tracktor['seed'])
torch.cuda.manual_seed(tracktor['seed'])
np.random.seed(tracktor['seed'])
torch.backends.cudnn.deterministic = True
output_dir = osp.join(get_output_dir(tracktor['module_name']),
tracktor['name'])
sacred_config = osp.join(output_dir, 'sacred_config.yaml')
if not osp.exists(output_dir):
os.makedirs(output_dir)
with open(sacred_config, 'w') as outfile:
yaml.dump(_config, outfile, default_flow_style=False)
##########################
# Initialize the modules #
##########################
# object detection
print("[*] Building object detector")
if tracktor['network'].startswith('frcnn'):
# FRCNN
from tracktor.frcnn import FRCNN
from frcnn.model import config
if _config['frcnn']['cfg_file']:
config.cfg_from_file(_config['frcnn']['cfg_file'])
if _config['frcnn']['set_cfgs']:
config.cfg_from_list(_config['frcnn']['set_cfgs'])
obj_detect = FRCNN(num_layers=101)
obj_detect.create_architecture(2,
tag='default',
anchor_scales=config.cfg.ANCHOR_SCALES,
anchor_ratios=config.cfg.ANCHOR_RATIOS)
obj_detect.load_state_dict(torch.load(tracktor['obj_detect_weights']))
else:
raise NotImplementedError(
f"Object detector type not known: {tracktor['network']}")
obj_detect.eval()
obj_detect.cuda()
# tracktor
tracker = Tracker(obj_detect, tracktor['tracker'])
tracker.reset() # init tracker
print("[*] Beginning evaluation...")
time_total = 0
cap = cv2.VideoCapture(webcam)
num_images = 0
images = []
try:
begin = time.time()
while (cap.isOpened()):
ret, frame = cap.read()
images.append(frame)
time.time()
try:
blob = data_handle.data_process(frame)
except:
print('over')
break
tracker.step(blob)
num_images += 1
if num_images % 10 == 0:
print('now is :', num_images)
results = tracker.get_results()
end = time.time()
print("[*] Tracks found: {}".format(len(results)))
print('It takes: {:.3f} s'.format((end - begin)))
if tracktor['write_images']:
plot_sequence(
results, images,
'/home/longshuz/project/tracking_wo_bnw/output/tracktor/results'
)
cap.release()
except:
raise KeyboardInterrupt