LiheYoung / Depth-Anything

[CVPR 2024] Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data. Foundation Model for Monocular Depth Estimation
https://depth-anything.github.io
Apache License 2.0
6.73k stars 516 forks source link

Fine-tune with custom dataset #35

Open kime541200 opened 7 months ago

kime541200 commented 7 months ago

This is really a fantastic project, thank you for your effort!

Will you provide methods to fine-tuning this model with custom dataset?

LiheYoung commented 7 months ago

We use ZoeDepth to fine-tune from the Depth Anything pre-trained encoder for metric depth estimation. You can simply follow this example.

Actually, if you already have your own codebase, all you need is to change your model to our Depth Anything model, i.e., using our pre-trained encoder and leaving the decoder randomly initialized, and then fine-tuned with your custom dataset.