search

antabangun / coex

GNU General Public License v3.0
142 stars 19 forks source link

Thank you for your great work! I would like to ask some questions #18

Open flyxuexi opened 1 month ago

flyxuexi commented 1 month ago

Thank you for your great work， I uploaded the disparity map in version_0/media generated through stereo.py to the kitti official website, and the index I got was slightly worse than the 3px (1.93) and D1(2.13) in the paper. What is the reason? In addition, I would like to ask you if I want to make some improvements to your model. What aspects can I start from? thank you very much！

antabangun commented 1 month ago

Thank you for your interest.

The slight discrepancy in performance metrics could stem from various factors, including differences in testing environments or hardware configurations. Without any details, I can't pinpoint the exact cause.

To improve the model, you might consider enhancing disparity prediction in occluded areas, reducing blurring around sharp depth edges, or improving the model in capturing thin structures, etc.

flyxuexi commented 1 month ago

Thank you for your interest.

The slight discrepancy in performance metrics could stem from various factors, including differences in testing environments or hardware configurations. Without any details, I can't pinpoint the exact cause.

To improve the model, you might consider enhancing disparity prediction in occluded areas, reducing blurring around sharp depth edges, or improving the model in capturing thin structures, etc.

thanks