Tobias-Fischer / rt_gene

RT-GENE: Real-Time Eye Gaze and Blink Estimation in Natural Environments
http://www.imperial.ac.uk/personal-robotics
Other
374 stars 71 forks source link

image type is .jpeg or .png #92

Closed lindayuanyuan closed 3 years ago

lindayuanyuan commented 3 years ago

When the im_path contains files with .jpeg or .png extensions, if I point the im_path to the folder, it seems that /rt_gene/rt_gene_standalone/estimate_gaze_standalone.py can't process those files. However, only if I point the im_path to the specific file, it works. What should I do if I want to process a batch of files with .jpeg or .png images? Thanks.

Tobias-Fischer commented 3 years ago

Hi, Thanks for this report! .png should work fine - I just committed https://github.com/Tobias-Fischer/rt_gene/commit/b4f582eafb56d1de6c361f815aaca268802c8222 which should also allow .png extension. Previously the CaPiTaLiZaTiOn mattered, whereas that commit fixes this, too.

Please let me know if your issue is addressed.

lindayuanyuan commented 3 years ago

Hi, Thanks for this report! .png should work fine - I just committed b4f582e which should also allow .png extension. Previously the CaPiTaLiZaTiOn mattered, whereas that commit fixes this, too.

Please let me know if your issue is addressed.

Thanks for the quick commit. Yes, it works now. When the image size is small, it throws errors like RuntimeError: Given input size: (512x1x2). Calculated output size: (512x0x1). Output size is too small, what is the size minimum requirement to process? thanks.

Tobias-Fischer commented 3 years ago

I'm not exactly sure - you can find the image sizes that we used in the paper. You can resize the images and do a trial and error investigation. It would also be good if you provide the full stack trace.

lindayuanyuan commented 3 years ago

resizing the image did the work. Thanks.