search

duanyiqun / DiffusionDepth

PyTorch Implementation of introducing diffusion approach to 3D depth perception
https://arxiv.org/abs/2303.05021
Apache License 2.0
293 stars 16 forks source link

For overfitting problem #39

Closed ZhengrongZhang closed 8 months ago

ZhengrongZhang commented 11 months ago

Sorry to bother you Thank you for the interesting work. During the training, I met overfitting problems. For the training process, RMSE and other indicates keep decreasing, however, for the val and testing process, the indicates keep changing irregularly. Can you give me some advice for such problem?

duanyiqun commented 11 months ago

May I ask what dataset are you using right now?

ZhengrongZhang commented 11 months ago

Sorry to replay you late. The dataset I use was Kitti

发自我的iPhone

------------------ Original ------------------ From: Yiqun Duan @.> Date: Thu,Oct 19,2023 6:33 PM To: duanyiqun/DiffusionDepth @.> Cc: ZhengrongZhang @.>, Author @.> Subject: Re: [duanyiqun/DiffusionDepth] For overfitting problem (Issue #39)

May I ask what dataset are you using right now?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

duanyiqun commented 11 months ago

For Kitti previous overfitting should be not that severe. A typical setting will be like training RMSE would be down to 0.9xx validation would be 1.3xx, and test would be 1.4xx. Maybe can adjust the data random cut width to 760-960. That will be more stable.