def run(model, val_loader, templates, prob_thresh, nms_thresh, device, split, debug=False):
for idx, (img, filename) in tqdm(enumerate(val_loader), total=len(val_loader)):
dets = trainer.get_detections(model, img, templates, val_loader.dataset.rf,
val_loader.dataset.transforms, prob_thresh,
nms_thresh, device=device)
write_results(dets, filename[0], split)
return dets
def run(model,
val_loader,
templates,
prob_thresh,
nms_thresh,
device,
split,
debug=False):
for idx, (img, filename) in tqdm(enumerate(val_loader),
total=len(val_loader)):
dets = trainer.get_detections(model,
img,
templates,
val_loader.dataset.rf,
val_loader.dataset.transforms,
prob_thresh,
nms_thresh,
device=device)
print("dets", dets.shape)
if debug:
print(img.shape)
mean = torch.as_tensor([0.485, 0.456, 0.406],
dtype=torch.float32,
device=img.device)
std = torch.as_tensor([0.229, 0.224, 0.225],
dtype=torch.float32,
device=img.device)
img = (img[0]).permute((1, 2, 0))
print(img.shape)
img = std * img + mean
img = np.clip(img.numpy(), 0, 1)
im = Image.fromarray((img * 255).astype('uint8'), 'RGB')
visualize.visualize_bboxes(im, dets)
write_results(dets, filename[0], split)
return dets
def run(model,
val_loader,
templates,
prob_thresh,
nms_thresh,
device,
split,
debug=False):
num_images = len(val_loader)
target_dets = np.zeros(shape=(num_images, 2, 100, 4))
target_scores = np.zeros(shape=(num_images, 2, 100, 2))
images = []
for idx, (img, filename) in enumerate(val_loader):
print("idx", idx)
dets = trainer.get_detections(model,
img,
templates,
val_loader.dataset.rf,
val_loader.dataset.transforms,
prob_thresh,
nms_thresh,
device=device)
#dets=dets[dets[:,-1].argsort()]
dets = dets[dets[:, -1] > 0.6]
print("number of dets:", dets.shape[0])
num_box = dets.shape[0]
target_det = dets[:, 0:4]
target_score = dets[:, -1]
target_dets[idx, 1, :num_box, :] = target_det
target_scores[idx, 1, :num_box, 1] = target_score
target_scores[idx, 1, :num_box, 0] = 1 - target_score
# if target_score.any():
# target_scores[idx,1,:num_box,1]=target_score
# target_scores[idx,1,:num_box,0]=np.max(1-target_score,0)
# else:
# target_scores[idx,1,:num_box,1]=0
# target_scores[idx,1,:num_box,0]=1
if idx != 0:
target_dets[idx, 0, :, :] = target_dets[idx - 1, 1, :, :]
target_scores[idx, 0, :, 1] = target_scores[idx - 1, 1, :, 1]
target_scores[idx, 0, :, 0] = target_scores[idx - 1, 1, :, 0]
else:
target_dets[idx, 0, :, :] = target_dets[idx, 1, :, :]
target_scores[idx, 0, :, 1] = target_scores[idx, 1, :, 0]
target_scores[idx, 0, :, 0] = target_scores[idx, 1, :, 1]
if debug:
print(img.shape)
mean = torch.as_tensor([0.485, 0.456, 0.406],
dtype=torch.float32,
device=img.device)
std = torch.as_tensor([0.229, 0.224, 0.225],
dtype=torch.float32,
device=img.device)
img = (img[0]).permute((1, 2, 0))
print(img.shape)
img = std * img + mean
img = np.clip(img.numpy(), 0, 1)
im = Image.fromarray((img * 255).astype('uint8'), 'RGB')
images.append(im)
#visualize.visualize_bboxes(im, dets)
#write_results(dets, filename[0], split)
return images, target_dets, target_scores