Hi, thanks for open-sourcing this excellent project.
I have been reading the paper and testing the released model recently. I noticed that, in some cases, the model produces noticeable depth jumps on non-Lambertian objects, which seems to affect its generalization in these challenging scenes.
Since the paper mentions the training setup, I wanted to ask whether you have plans to release the training scripts as well. I am interested in fine-tuning the model on some specific datasets to explore whether its performance on such scenes can be improved.
If possible, could you also share whether there is an estimated timeline for this release?
Thanks again for your great work.
Hi, thanks for open-sourcing this excellent project.
I have been reading the paper and testing the released model recently. I noticed that, in some cases, the model produces noticeable depth jumps on non-Lambertian objects, which seems to affect its generalization in these challenging scenes.
Since the paper mentions the training setup, I wanted to ask whether you have plans to release the training scripts as well. I am interested in fine-tuning the model on some specific datasets to explore whether its performance on such scenes can be improved.
If possible, could you also share whether there is an estimated timeline for this release?
Thanks again for your great work.
