Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix mask for multi object persam (persam_f_multi_obj.py) #54

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix mask for multi object persam (persam_f_multi_obj.py) #54

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 14 additions & 14 deletions persam_f_multi_obj.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@


def get_arguments():

parser = argparse.ArgumentParser()

parser.add_argument('--data', type=str, default='./data')
Expand All @@ -30,10 +30,10 @@ def get_arguments():
parser.add_argument('--train_epoch_inside', type=int, default=200)
parser.add_argument('--log_epoch', type=int, default=200)
parser.add_argument('--training_percentage', type=float, default=0.5)

parser.add_argument('--max_objects', type=int, default=10)
parser.add_argument('--iou_threshold', type=float, default=0.8)

args = parser.parse_args()
return args

Expand All @@ -47,11 +47,11 @@ def main():
masks_path = args.data + '/Annotations/'
output_path = './outputs/' + args.outdir



if not os.path.exists('./outputs/'):
os.mkdir('./outputs/')

for obj_name in os.listdir(images_path):
persam_f(args, obj_name, images_path, masks_path, output_path)

Expand All @@ -66,8 +66,8 @@ def persam_f(args, obj_name, images_path, masks_path, output_path):
device = "cuda" if torch.cuda.is_available() else "cpu"
sam = sam_model_registry[sam_type](checkpoint=sam_ckpt).to(device=device)
sam.eval()


for name, param in sam.named_parameters():
param.requires_grad = False
predictor = SamPredictor(sam)
Expand All @@ -90,9 +90,9 @@ def persam_f(args, obj_name, images_path, masks_path, output_path):
ref_mask = cv2.imread(ref_mask_path)
ref_mask = cv2.cvtColor(ref_mask, cv2.COLOR_BGR2RGB)

gt_mask = torch.tensor(ref_mask)[:, :, 0] > 0
gt_mask = torch.tensor(ref_mask)[:, :, 0] > 0
gt_mask = gt_mask.float().unsqueeze(0).flatten(1).cuda()

# print("======> Obtain Self Location Prior" )
# Image features encoding
ref_mask = predictor.set_image(ref_image, ref_mask)
Expand Down Expand Up @@ -129,7 +129,7 @@ def persam_f(args, obj_name, images_path, masks_path, output_path):
# Learnable mask weights
mask_weights = Mask_Weights().cuda()
mask_weights.train()

optimizer = torch.optim.AdamW(mask_weights.parameters(), lr=args.lr, eps=1e-4)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, args.train_epoch_inside)

Expand Down Expand Up @@ -175,7 +175,7 @@ def persam_f(args, obj_name, images_path, masks_path, output_path):
test_image = cv2.cvtColor(test_image, cv2.COLOR_BGR2RGB)
test_image_original = cv2.imread(test_image_path)
test_image_original = cv2.cvtColor(test_image_original, cv2.COLOR_BGR2RGB)

history_masks = []
plt.figure(figsize=(10, 10))
for i in tqdm(range(args.max_objects)):
Expand Down Expand Up @@ -261,14 +261,14 @@ def persam_f(args, obj_name, images_path, masks_path, output_path):
show_points(topk_xy, topk_label, plt.gca())
history_masks.append(mask_colors)
# Save masks

plt.imshow(test_image_original)
vis_mask_output_path = os.path.join(output_path, f'vis_mask_{test_idx}_objects:{len(history_masks)}.jpg')
with open(vis_mask_output_path, 'wb') as outfile:
plt.savefig(outfile, format='jpg')

mask_output_path = os.path.join(output_path, test_idx + '.png')
cv2.imwrite(mask_output_path, mask_colors)
cv2.imwrite(mask_output_path, sum(history_masks))



Expand All @@ -287,7 +287,7 @@ def point_selection(mask_sim, topk=1):
topk_xy = torch.cat((topk_y, topk_x), dim=0).permute(1, 0)
topk_label = np.array([1] * topk)
topk_xy = topk_xy.cpu().numpy()

return topk_xy, topk_label


Expand Down