Training Hyperparameter to Reproduce Paper Result

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

Training Hyperparameter to Reproduce Paper Result #21

Closed andrewivan123 closed 4 months ago

andrewivan123 commented 6 months ago

I would like to reproduce the paper results so what are the training hyperparameters used in the paper (e.g batch size, no. of epochs, loss parameters, etc)?

zhyever commented 4 months ago

All the hyperparameters are given in the corresponding config files