search

zhyever / PatchFusion

[CVPR 2024] An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation
https://zhyever.github.io/patchfusion/
MIT License
926 stars 62 forks source link

About trainnig weights and inference #5

Closed Hank10032 closed 7 months ago

Hank10032 commented 7 months ago

Your work is excellent. I would like to inquire about when the trained weights will be released. I tried to infer some images using the method set in GitHub, but it shows that the keys of the model and weights do not match.

zhyever commented 7 months ago

Thanks for the interest in our work! I have update the download link. Please check it again. If the weights are still mismatched, please send the error information. Thank you for your help

Hank10032 commented 7 months ago

Thank you for your patience in answering me.But I still have the following problem when using infer. This is the code I use. python ./infer_user.py --model zoedepth_custom --ckp_path /media/TB2/PatchFusion/patchfusion_u4k.pt --model_cfg_path ./zoedepth/models/zoedepth_custom/configs/config_zoedepth_patchfusion.json --rgb_dir /media/TB2/Processed/ADE_20K/image --show --show_path /media/TB2/Processed/ADE_20K/show --save --save_path /media/TB2/Processed/ADE_20K/depth_patch --mode r128 --boundary 0 --blur_mask This is the problem image Thank you again for your patience

zhyever commented 7 months ago

aha, I get it. Please check your timm version. You should have used a higher version timm than 0.6.12. Please downgrade to 0.6.12.

Hank10032 commented 7 months ago

Thank you for your high-quality and patient answers. All my questions have been solved. Do you have any plans to release batch inference code?