Regarding the test results of ETH3D

[cangmis]

When I used the ETH3D preprocessed dataset provided by PatchMatchNet for measurement, the generated point cloud was extremely sparse. What could be the reason? Is it because of the dataset? Could you provide the ETH3D dataset that you used during testing?

[susuwj]

It seems to be caused by the incorrect depth range. The last line of the cams.txt provided by PathmatchNet is depth_min depth_max, while the last line of the cams.txt we used is depth_min depth_interval depth_num depth_max. You need to modify the function read_cam_file() in datasets/eth_ eval.py to read the correct depth range for MVS.

[cangmis]

Thank you very much for your reply. I used the colmap2mvsnet.py file in the CasMVSNet network to generate the poses of eth3d. However, it seems that there is a deviation in the position of the generated point cloud, resulting in zero accuracy and completeness for the evaluation results. So could you please provide the corresponding scripts for generating poses? 1228711396 ***@***.***  

…

------------------ 原始邮件 ------------------ 发件人: "susuwj/EPNet" ***@***.***>; 发送时间: 2024年11月13日(星期三) 下午4:03 ***@***.***>; ***@***.******@***.***>; 主题: Re: [susuwj/EPNet] Regarding the test results of ETH3D (Issue #2) It seems to caused by the incorrect depth range. The last line of the cams.txt provided by PathmatchNet is depth_min depth_max, while the last line of the cams.txt we used is depth_min depth_interval depth_num depth_max. You need to modify the function read_cam_file() in datasets/eth_ eval.py to read the correct depth range for MVS. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>