Open HaosenZ opened 4 months ago
If you finetune the model against gt depth maps while following https://github.com/DepthAnything/Depth-Anything-V2/issues/102#issuecomment-2275132770 results should be depth predictions (furthermore metric_depth/train.py is designed to fine-tune architecture wrt gt depth maps)
Thank you for your contribution I saw in your metric depth description that the output of the pre trained model can be used as a disparity map. Now, I want to use a custom dataset that includes RGB images and a 16 bit depth map to fine tune the outdoor model. Is the resulting output a depth map?