search

imatge-upc / rvos

RVOS: End-to-End Recurrent Network for Video Object Segmentation (CVPR 2019)
https://imatge-upc.github.io/rvos/
Other
277 stars 54 forks source link

To run my own videos error #10

Closed zdwong closed 5 years ago

zdwong commented 5 years ago

I use this command to run my own video: python demo.py -model_name one-shot-model-davis -frames_path path-to-your-frames --zero_shot --overlay_masks

Howerver, it appears an error: image

agirbau commented 5 years ago

I think you are using a one-shot model (which requires an initial input mask) and using a zero-shot configuration. Try using zero-shot-model-davis and check if the problem persists.

agirbau commented 5 years ago

Btw pull to the lastest version to to run it.

zdwong commented 5 years ago

I have use the latest version: git clone https://github.com/imatge-upc/rvos.git. If I use zero-shot-model-davis, the problems still exists. @agirbau

agirbau commented 5 years ago

Can you copy and paste your cmd output here?

zdwong commented 5 years ago

The same problem. I have tried it before. @agirbau

agirbau commented 5 years ago

Copy and paste your command line prompt in next comment and I'll check :)

zdwong commented 5 years ago

ok, wait for a while. I have to reconfigure the running environment again. @agirbau I can make sure that the problem is the same.

zdwong commented 5 years ago

When I run the command again, it doesn't show the error now. But the video has no segmentation result.

python demo.py -model_name zero-shot-model-davis -frames_path /detection/data/ch00011_frames --zero_shot --overlay_masks

image

When I run the comand of this: python demo.py -model_name one-shot-model-davis --overlay_masks , it has the segmentation result. @agirbau

agirbau commented 5 years ago

That's weird. Can you double check that you don't have the results in models/zero-shot-model-davis/results?

zdwong commented 5 years ago

I use my own video data, not the DAVIS. The path has the result. But every frame doesn't have the segmentation result. I have check two times. @agirbau

zdwong commented 5 years ago

I run the DAVIS video again, but it has no segmentation result too. image The result picture: image

if I run the command: python demo.py -model_name one-shot-model-davis --overlay_masks The result picture: image @agirbau

agirbau commented 5 years ago

I will check next week then!

agirbau commented 5 years ago

Ok, do a pull and check it again.

zdwong commented 5 years ago

Yeah, I have tried the new version, It has the segmentation now. But the segmentation result cannot meet my expectation.

carlesventura commented 5 years ago

As you can see in the Tables of the paper the model proposed works much better in the one-shot (or semi-supervised) scenario than in the zero-shot (or unsupervised) scenario. We are working to improve our results for the zero-shot case.

zdwong commented 5 years ago

Thanks for your great work trying to do zero-shot(or unsupervised) scenario. I think it's an valuable and interesting research topic. @agilmor @carlesventura

melvin0kuo commented 11 months ago

where can find demo.py？didnt find it ：（