NVIDIA / flownet2-pytorch

Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks
Other
3.09k stars 739 forks source link

I follow the inference instruction in README.md,but I still get blurry result. #228

Open coly-ai opened 3 years ago

coly-ai commented 3 years ago

Here is my instruction:I just run it~

CUDA_VISIBLE_DEVICES=1 python main.py --inference --model FlowNet2 --save_flow \
--inference_dataset ImagesFromFolder \
--inference_dataset_root /home/xxx/xxxxx/file001/ \
--resume /home/xxx/flownet2-pytorch/checkpoint/FlowNet2_checkpoint.pth.tar \
--save output/

Then I got the result on my dataset which the flow look so blurry. 000000 flo 000001 flo I believe there is something wrong,the background is still~ I am very greatful if you could tell me how I can improve the performance!

wgh-2018 commented 2 years ago

I encounter the same issue,have you solved the problem?