how to work with fixed region inpainting

[IamMegatron2025]

hello, how to work with fixed region inpainting, such as the logo with the fixed position on the video. I follow your instruction:

CUDA_VISIBLE_DEVICES=0 python tools/video_inpaint.py --frame_dir ./demo/lady-running/frames \
--MASK_ROOT ./demo/lady-running/mask_bbox.png \
--img_size 448 896 --DFC --FlowNet2 --Propagation \
--PRETRAINED_MODEL_1 ./pretrained_models/resnet50_stage1.pth \
--PRETRAINED_MODEL_2 ./pretrained_models/DAVIS_model/davis_stage2.pth \
--PRETRAINED_MODEL_3 ./pretrained_models/DAVIS_model/davis_stage3.pth \
--MS --th_warp 3 --FIX_MAS

But it still does Iter for many minutes. Why it so slow...

[nbei]

I provide several ways for reducing running time:

• using single stage model
• get th_warp much higher like 10
• decrease the resolution of input

However, these tricks will harm the final performance. The running speed depends on many things, like CPUs, memory bus. I'm still working on the more efficient versioin and welcome to any kind of suggestions.