def runQuery(self):
self.queryTrajectory = GeographicTrajectory(self.trajectoryBag)
# XXX: query trajectory should be made on specified time range
samples = self.imageBag.desample(-1,
True,
startOffsetTime=self.startOffset,
stopOffsetTime=self.stopOffset)
print("Ready")
pool4 = mlc.SuperPool(n_cpu=self.numCPU)
positions = pool4.map(self.processQuery, samples)
print("Done")
return positions
def main():
threshold = 0.5
crop_version = 'cropv3'
crop_size = 512
np.random.seed(100)
pool = mlc.SuperPool(16)
for dataset in ['CheXpert-v1.0']:
print(dataset)
mask_dir = opj(DATA_DIR, 'external', dataset, 'images/masks_1024')
image_dir = opj(DATA_DIR, 'external', dataset, 'images/images_1024')
out_mask_dir = opj(DATA_DIR, 'external', dataset,
'images/masks_%s' % (crop_version))
out_image_dir = opj(DATA_DIR, 'external', dataset,
'images/images_%s' % (crop_version))
os.makedirs(out_mask_dir, exist_ok=True)
os.makedirs(out_image_dir, exist_ok=True)
result_dir = opj(
RESULT_DIR, 'submissions',
'siim_lung_nobalance_unet_resnet34_cbam_v0a_i256_aug9_symmetriclovaszloss_4folds',
'fold0/epoch_final/default')
result_df = pd.read_csv(
opj(result_dir, 'results_%s.csv.gz' % 'chexpert'))
result_probs = np.load(opj(result_dir,
'prob_%s.npz' % 'chexpert'))['data']
print(result_df.shape, result_probs.shape)
params = []
for idx, fname in tqdm(enumerate(result_df[ID].values),
total=len(result_df)):
param = (idx, crop_size, fname, result_probs[idx], threshold,
image_dir, mask_dir, out_mask_dir, out_image_dir, dataset)
params.append(param)
result_list = pool.map(generate_crop_images,
params,
description='create cropv3')
box_df = pd.concat(result_list, ignore_index=True)
box_file = opj(DATA_DIR, 'external', dataset,
'images/boxes_%s.csv.gz' % (crop_version))
box_df.to_csv(box_file, index=False, compression='gzip')
train_df = get_train_df_ohe()
# external images
public_df = get_public_df_ohe()
colors = ['red', 'green', 'blue']
train_df = generate_hash(train_df,
colors,
dataset='train',
is_update=False)
public_df = generate_hash(public_df,
colors,
dataset='external',
is_update=False)
pool = mlc.SuperPool(3)
params = []
for color in colors:
base_tran_hash = train_df[color].values
base_external_hash = public_df[color].values
train_ids = train_df['img_base_path'].values
external_ids = public_df['img_base_path'].values
base_hash_all = np.concatenate((base_tran_hash, base_external_hash))
ids_all = np.concatenate((train_ids, external_ids))
params.append(
(color, threshold, base_hash_all, base_hash_all, ids_all, ids_all))
df_list = pool.map(calc_hash, params)
src, fname, dst, size = param
color = fname.replace('.jpg', '').split('_')[-1]
try:
im = np.array(Image.open(opj(src, fname)))[:, :,
COLOR_INDEXS.get(color)]
except:
print('bad image : %s' % fname)
im = cv2.imread(opj(src, fname))[:, :, -1::-1][:, :,
COLOR_INDEXS.get(color)]
im = cv2.resize(im, (size, size), interpolation=cv2.INTER_LINEAR)
cv2.imwrite(opj(dst, fname), im, [int(cv2.IMWRITE_JPEG_QUALITY), 85])
if __name__ == '__main__':
print('%s: calling main function ... ' % os.path.basename(__file__))
size = args.size
src = opj(EXTERNEL_DIR, 'HPAv18/jpg')
dst = opj(DATA_DIR, 'train/external_v18_%d' % size)
fnames = np.sort(os.listdir(src))
os.makedirs(dst, exist_ok=True)
start_num = max(0, len(os.listdir(dst)) - 48)
fnames = fnames[start_num:]
params = [(src, fname, dst, size) for fname in fnames]
pool = mlc.SuperPool(10)
pool.map(do_resize, params, description='resize image')
print('success.')
parser.add_argument("--ime",
type=str,
choices=enhanceMethods,
help='Preprocess image with this method')
else:
print("Enhancement not available; install im_enhance if you want")
cmdArgs = parser.parse_args()
queryBag = ImageBag(cmdArgs.bagfile, cmdArgs.topic)
mapsource = None
mapsource = GenericImageDatabase(cmdArgs.mapfile)
if hasattr(cmdArgs, 'ime') and (cmdArgs.ime is not None):
from place_recognizer.GenericImageMap import ime
mapsource.useEnhancement = True
mapsource.enhanceMethod = eval('ime.' + prog_arguments.ime)
else:
imeMethod = False
bagLock = Lock()
orb = cv2.ORB_create(6000)
print("Ready")
pool4 = mlc.SuperPool(n_cpu=1)
samples = queryBag.desample(-1, True, cmdArgs.start, cmdArgs.stop)
positions = pool4.map(processQuery, samples)
mlc.save(positions, cmdArgs.output)
print("Done")
def do_convert(fname_tif):
img = np.array(Image.open(opj(tif_dir, fname_tif)))
img = cv2.resize(img, (size, size), interpolation=cv2.INTER_LINEAR)
fname_img = fname_tif.replace('.tif', '.png')
cv2.imwrite(opj(img_dir, fname_img), img)
parser = argparse.ArgumentParser(description='PyTorch Protein Classification')
parser.add_argument('--dataset', type=str, default='train', help='dataset')
parser.add_argument('--size', type=int, default=1536, help='size')
args = parser.parse_args()
if __name__ == '__main__':
print('%s: calling main function ... ' % os.path.basename(__file__))
size = args.size
dataset = args.dataset
tif_dir = opj(TIF_DIR, dataset, 'tifs')
img_dir = opj(DATA_DIR, dataset, 'images_%d' % size)
n_cpu = 3
os.makedirs(img_dir, exist_ok=True)
start_num = max(0, len(os.listdir(img_dir)) - n_cpu * 2)
fnames = np.sort(os.listdir(tif_dir))[start_num:]
pool = mlc.SuperPool(n_cpu)
df_list = pool.map(do_convert,
fnames,
description='resize %s image' % dataset)
print('\nsuccess!')