def detect(opt, logger, cfg):
demo = FeatureExtractionDemo(cfg, parallel=opt.parallel)
query_loader = []
print(len(os.listdir(opt.input_picture_path)))
for query_image in os.listdir(opt.input_picture_path):
query_image = cv2.imread(
os.path.join(opt.input_picture_path, query_image))
query_image = cv2.cvtColor(query_image,
cv2.COLOR_BGR2RGB) # PIL: (233, 602)
query_image = cv2.resize(query_image, (128, 256))
query_image = np.transpose(query_image, (2, 0, 1))
query_feats = demo.run_on_image(
torch.from_numpy(query_image).unsqueeze(0))
print(query_feats.shape)
query_loader.append(query_feats)
extract_features = torch.cat(query_loader, dim=0).data.cpu().numpy()
# extract_features = torch.nn.functional.normalize(query_feats, dim=1, p=2).data.cpu().numpy()
print('features:', extract_features.shape)
with Timer('All Steps'):
global args
# args = parser.parse_args()
label_path = None
pred_label_path = None
print('=> Use cuda ?: {}'.format(opt.is_cuda))
# with Timer('Extract Feature'):
# extract_features = extract_fature(args)
if eval(opt.is_evaluate):
opt.label_path = 'data/tmp/test.meta'
if not eval(opt.is_cuda):
opt.knn_method = 'faiss-cpu'
with Timer('Face Cluster'):
cluster_main(opt, extract_features)
print(
"=> Face cluster done! The cluster results have been saved in {}".
format(opt.output_picture_path))
class Reid_feature():
def __init__(self):
args = get_parser().parse_args()
cfg = setup_cfg(args)
self.demo = FeatureExtractionDemo(cfg, parallel=args.parallel)
def __call__(self, img_list):
import time
t1 = time.time()
feat = self.demo.run_on_image(img_list)
# print('reid time:', time.time() - t1, len(img_list))
return feat
help='path to save features')
parser.add_argument(
"--opts",
help="Modify config options using the command-line 'KEY VALUE' pairs",
default=[],
nargs=argparse.REMAINDER,
)
return parser
if __name__ == '__main__':
args = get_parser().parse_args()
cfg = setup_cfg(args)
demo = FeatureExtractionDemo(cfg,
device=args.device,
parallel=args.parallel)
PathManager.mkdirs(args.output)
if args.input:
if PathManager.isdir(args.input[0]):
args.input = glob.glob(os.path.expanduser(args.input[0]))
assert args.input, "The input path(s) was not found"
for path in tqdm.tqdm(args.input):
img = cv2.imread(path)
feat = demo.run_on_image(img)
feat = feat.numpy()
np.save(
os.path.join(args.output,
path.replace('.jpg', '.npy').split('/')[-1]),
feat)
frame_id = 1
ret = True
if vidcap.isOpened():
while ret:
ret, frame = vidcap.read()
if frame_id in frames:
bboxes = det_results[str(frame_id)]
save_results[str(frame_id)] = []
for bbox in bboxes:
box = bbox[0]
pid = bbox[1]
box_i = [int(x) for x in box]
x, y, w, h = box_i
img = frame[y:y + h, x:x + w:, ]
# img = img.to(device)
feat = demo.run_on_image(img)[0].numpy().tolist()
save_results[str(frame_id)].append([box, pid, feat])
frame_id += 1
if frame_id > max_frame:
break
#save_json(osp.join(args.output, args.det_type, videoname + '.final.reduced.json'), save_results)
# cores = args.cores
# if cores <= 0:
# cores = cpu_count()
# num_of_process = int(min(cores, len(videos)))
# mp_pool = Pool(cores)
# multi_inputs = []
# for i in range(num_of_process):
# param = (videos[i], i)
# multi_inputs.append(param)