Open lovejoker1102 opened 2 years ago
你好! 我也是遇到关于colmap生成参数有问题的情况 我将提供的测试集的图用colmap+colmap2mvsnet.py跑出来的结果与实际提供的差很多 当然生成的结果也不对 但不知道错在哪 请问大佬解决了吗
你好! 我也是遇到关于colmap生成参数有问题的情况 我将提供的测试集的图用colmap+colmap2mvsnet.py跑出来的结果与实际提供的差很多 当然生成的结果也不对 但不知道错在哪 请问大佬解决了吗
mvsnet的相机参数我记得是从ground truth转换过来的,colmap跑出来的精度没有那么高,所以结果要比原来差
大佬知道怎么转换吗 ground truth是指啥 标定得到的相机参数?
对的,dtu有标定好的相机参数Sent from my iPhone------------------ Original ------------------From: hhl134 @.>Date: Tue,Jun 20,2023 7:48 PMTo: YoYo000/MVSNet @.>Cc: KissSheep @.>, Author @.>Subject: Re: [YoYo000/MVSNet] camera parameters from colmap have some problems(Issue #147)
大佬知道怎么转换吗 ground truth是指啥 标定得到的相机参数?
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: @.> [ { @.": "http://schema.org", @.": "EmailMessage", "potentialAction": { @.": "ViewAction", "target": "https://github.com/YoYo000/MVSNet/issues/147#issuecomment-1598625122", "url": "https://github.com/YoYo000/MVSNet/issues/147#issuecomment-1598625122", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { @.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]
现在我是想自己做测试集 他也提供了colmap2mvsnet.py 就大佬上面问的问题 想问下大佬解决了吗 就colmap生成出来的相机参数不能用的情况 我不管是用dtu还是tat哪个数据集来用colmap生成都不行 都不是他给的数据的效果
你说的是参数跟他的参数看起来差别大还是说跑出来效果差别大Sent from my iPhone------------------ Original ------------------From: hhl134 @.>Date: Tue,Jun 20,2023 8:27 PMTo: YoYo000/MVSNet @.>Cc: KissSheep @.>, Author @.>Subject: Re: [YoYo000/MVSNet] camera parameters from colmap have some problems(Issue #147) 现在我是想自己做测试集 他也提供了colmap2mvsnet.py 就大佬上面问的问题 想问下大佬解决了吗 就colmap生成出来的相机参数不能用的情况 我不管是用dtu还是tat哪个数据集来用colmap生成都不行 都不是他给的数据的效果
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: @.> [ { @.": "http://schema.org", @.": "EmailMessage", "potentialAction": { @.": "ViewAction", "target": "https://github.com/YoYo000/MVSNet/issues/147#issuecomment-1598677480", "url": "https://github.com/YoYo000/MVSNet/issues/147#issuecomment-1598677480", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { @.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]
两个都是有 首先就是相机的外参跑出来就差很多 对应的深度值就不对了 那生成出来的深度图 点云图肯定就不对了
Colman跑出来确实有误差 效果比原来差是正常的 相机参数差很多是因为坐标系选择不一样 深度不一样也是尺度不一样导致的 原因还是出在colmap精度不够发自我的iPhone------------------ Original ------------------From: hhl134 @.>Date: Tue,Jun 20,2023 8:40 PMTo: YoYo000/MVSNet @.>Cc: KissSheep @.>, Author @.>Subject: Re: [YoYo000/MVSNet] camera parameters from colmap have some problems(Issue #147) 两个都是有 首先就是相机的外参跑出来就差很多 对应的深度值就不对了 那生成出来的深度图 点云图肯定就不对了
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: @.> [ { @.": "http://schema.org", @.": "EmailMessage", "potentialAction": { @.": "ViewAction", "target": "https://github.com/YoYo000/MVSNet/issues/147#issuecomment-1598695130", "url": "https://github.com/YoYo000/MVSNet/issues/147#issuecomment-1598695130", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { @.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]
@HaoyangHuang Hi, I also have this problem now and could you please provide more specific information on the different coordinates used from "相机参数差很多是因为坐标系选择不一样". Thanks a lot~
Hi, I recently try to run colmap to get the camera parameters, but I found something confused. Colmap's input is DTU scans, when feature extraction I use PINHOLE camera model which is same as yours. However, after I converting the colmap output to mvsnet input by
colmap2mvsnet.py
, I have found the camera intrisic and extrinsic parameters are quite different from yours. Could you tell me how you run colmap (maybe some setting I have ignored?)