search

kaichen-z / Manydepth2

Manydepth2: Motion-Aware Self-Supervised Monocular Depth Estimation in Dynamic Scenes
https://arxiv.org/pdf/2312.15268
103 stars 17 forks source link

Solution of synthetic ref images without interference from dynamic object #2

Closed Qianqian3764 closed 1 month ago

Qianqian3764 commented 1 month ago

thanks for your excellent job!

I have a question about synthetic ref images：new_static_reference = ref_image*(~seg_ref) + static_reference_dyn*seg_ref However, I think this is still the original reference image, no changed, because the reconstruction of the dynamic object in the reference image is obtained using optical flow. In Dynamic depth(2022 ECCV), the reconstruction of the dynamic target in the reference image is obtained by using the depth and pose, that is, the position of the target in the target image under the assumption of static situation. The position of the dynamic target is changed in the obtained synthetic ref image.

kaichen-z commented 1 month ago

Dear Qianqian,

Thank you very much for your detailed review. You are correct, and we mistakenly uploaded the wrong version of our code. Please find the updated version.

kaichen-z commented 1 month ago

Please let me know if there are any further issues. If not, we will proceed to close this comment.