Question about keep my own image size

[youyoubud]

I'd like to visualize an image sequence extracted from a video, and I really need the output .flo file size to be consistent with my input image. How can I do that?
My image is (120, 180, 3).
Thank you!

[drinkingcoder]

We provide an option `keep_size' here. You can enable it for visualization.

[youyoubud]

I have tried this, but it didn't work.
computing flow... Traceback (most recent call last): File "visualize_flow.py", line 213, in <module> visualize_flow(root_dir, viz_root_dir, model, img_pairs, args.keep_size) File "visualize_flow.py", line 157, in visualize_flow flow, weights = compute_flow(model, image1, image2, weights) File "visualize_flow.py", line 76, in compute_flow weights = compute_weight(hws, image_size, sigma=0.05) File "visualize_flow.py", line 61, in compute_weight weights[:, idx, h:h+patch_size[0], w:w+patch_size[1]] = weights_hw RuntimeError: The expanded size of the tensor (180) must match the existing size (960) at non-singleton dimension 2. Target sizes: [1, 120, 180]. Tensor sizes: [432, 960]
Did I do something wrong?
Thank you again!

[drinkingcoder]

I updated the visualize_flow.py. You can try again.

[youyoubud]

It works, thank you!