def embedding(self, im, bbox):
im_in = im
cords = bbox[0:4]
center, scale = self.x1y1x2y2_to_cs(cords)
r = 0
trans = get_affine_transform(center[0], scale[0], r, (288, 384))
input_image = cv2.warpAffine(im_in,
trans, (288, 384),
flags=cv2.INTER_LINEAR)
with torch.no_grad():
adj = self.reset_adj_no1()
adj = adj.unsqueeze(0)
adj = adj.type(torch.cuda.FloatTensor)
input = self.transform(input_image)
input = input.unsqueeze(0)
with torch.cuda.device(self.gpu_id):
input = input.cuda()
feature = self.model(input, adj, flag=1)
if cfg.TEST.FLIP_TEST:
input_flipped = np.flip(input.cpu().numpy(), 3).copy()
input_flipped = torch.from_numpy(input_flipped).cuda()
feature_flipped = self.model(input_flipped, adj, flag=1)
feature = (feature + feature_flipped) * 0.5
feature = feature.cpu().numpy().squeeze()
return feature.tolist()
def extract_feature(self, im, bbox):
im_in = im
cords = bbox[0:4]
center, scale = self.x1y1x2y2_to_cs(cords)
r = 0
trans = get_affine_transform(center[0], scale[0], r, self.image_size)
input_image = cv2.warpAffine(
im_in,
trans, (int(self.image_width), int(self.image_height)),
flags=cv2.INTER_LINEAR)
with torch.no_grad():
input = self.transform(input_image)
input = input.unsqueeze(0)
with torch.cuda.device(self.gpu_id):
input = input.cuda()
feature = self.model(input, flag=1)
if cfg.TEST.FLIP_TEST:
input_flipped = np.flip(input.cpu().numpy(), 3).copy()
input_flipped = torch.from_numpy(input_flipped).cuda()
feature_flipped = self.model(input_flipped, flag=1)
feature = (feature + feature_flipped) * 0.5
feature = feature.cpu().numpy().squeeze()
return feature
def detect_pose(self, im, bbox): im_in = im cords = bbox[0:4] center, scale = self.x1y1x2y2_to_cs(cords) r=0 trans = get_affine_transform(center[0], scale[0], r, self.image_size) input_image = cv2.warpAffine(im_in, trans, (int(self.image_width), int(self.image_height)), flags=cv2.INTER_LINEAR) with torch.no_grad(): input = self.transform(input_image) input = input.unsqueeze(0) with torch.cuda.device(self.gpu_id): input = input.cuda() output = self.model(input) if cfg.TEST.FLIP_TEST: input_flipped = np.flip(input.cpu().numpy(),3).copy() input_flipped = torch.from_numpy(input_flipped).cuda() output_flipped = self.model(input_flipped) output_flipped = flip_back(output_flipped.cpu().numpy(), self.flip_pairs) output_flipped = torch.from_numpy(output_flipped.copy()).cuda() if cfg.TEST.SHIFT_HEATMAP: output_flipped[:, :, :, 1:] = output_flipped.clone()[:, :, :, 0:-1] output = (output + output_flipped) *0.5 output = output.cpu().numpy() preds, maxvals = self.get_final_preds(output, center, scale) preds, maxvals = preds.squeeze(), maxvals.squeeze() heatmaps = output.squeeze() #For posetrack dataset, the 3th and 4th channel is None preds = np.delete(preds, [3,4], axis=0) maxvals = np.delete(maxvals, [3,4], axis=0) heatmaps = np.delete(heatmaps, [3,4], axis=0) #print(heatmaps.shape,preds,maxvals) return preds, maxvals, heatmaps
def __getitem__(self, idx):
db_rec = copy.deepcopy(self.db[idx])
image_file = db_rec['image']
im_id = db_rec['image_id'] if 'image_id' in db_rec else 0
im_bbox = np.array(db_rec['bbox']) if 'bbox' in db_rec else np.array(
[0, 0, 0, 0])
#print(im_bbox)
filename = db_rec['filename'] if 'filename' in db_rec else ''
imgnum = db_rec['imgnum'] if 'imgnum' in db_rec else ''
if self.data_format == 'zip':
from utils import zipreader
data_numpy = zipreader.imread(
image_file, cv2.IMREAD_COLOR | cv2.IMREAD_IGNORE_ORIENTATION)
else:
data_numpy = cv2.imread(
image_file, cv2.IMREAD_COLOR | cv2.IMREAD_IGNORE_ORIENTATION)
joints = db_rec['joints_3d']
joints_vis = db_rec['joints_3d_vis']
c = db_rec['center']
s = db_rec['scale']
score = db_rec['score'] if 'score' in db_rec else 1
r = 0
if self.is_train:
sf = self.scale_factor
rf = self.rotation_factor
s = s * np.clip(np.random.randn() * sf + 1, 1 - sf, 1 + sf)
r = np.clip(np.random.randn()*rf, -rf*2, rf*2) \
if random.random() <= 0.6 else 0
if self.flip and random.random() <= 0.5:
data_numpy = data_numpy[:, ::-1, :]
joints, joints_vis = fliplr_joints(joints, joints_vis,
data_numpy.shape[1],
self.flip_pairs)
c[0] = data_numpy.shape[1] - c[0] - 1
trans = get_affine_transform(c, s, r, self.image_size)
input = cv2.warpAffine(
data_numpy,
trans, (int(self.image_size[0]), int(self.image_size[1])),
flags=cv2.INTER_LINEAR)
if self.transform:
input = self.transform(input)
for i in range(self.num_joints):
if joints_vis[i, 0] > 0.0:
joints[i, 0:2] = affine_transform(joints[i, 0:2], trans)
target, target_weight = self.generate_target(joints, joints_vis)
target = torch.from_numpy(target)
target_weight = torch.from_numpy(target_weight)
meta = {
'image': image_file,
'image_id': im_id,
'im_bbox': im_bbox,
'filename': filename,
'imgnum': imgnum,
'joints': joints,
'joints_vis': joints_vis,
'center': c,
'scale': s,
'rotation': r,
'score': score
}
# print(meta)
return input, target, target_weight, meta