feat(ml): inference_vid with canny

[wr0124]

Video Inference Debugging
1 With Canny Edge Detection:

• First Frame: Apply Canny edge detection.
• Subsequent Frames: Use the same noise mask as the first frame.

2 Without Canny Edge Detection:

• All Frames: Use the same noise mask throughout the sequence.

cd scripts/
mkdir -p ../inference_mario_bz
python3 -W ignore::UserWarning gen_vid_diffusion.py
--model_in_file /path/to/checkpoints/test_mario_vid_example_canny_drop2/latest_net_G_A.pth
--img_in /path/to/paths_part.txt
--paths_in_file /path/to/online_mario2sonic_full_mario/trainA/paths.txt
--mask_in /paths_part.txt
--dir_out ../inference_mario_bz
--sampling_steps 1000
--gpuid 2
~