MVSNet_pytorch
MVSNet_pytorch copied to clipboard
xy-guo
Terrible reconstruction effect on TankandTemples dataset
pre-trained model cannot reconstruct on other dataset, such as TankandTemples.
I resized the images from TankandTemples to 1600X1200 but the result is not good.
All of those are like this
@wangchengze001 I also encountered the same problem, how did you solve this problem in the end?
@YuhsiHu @xy-guo May I ask how to test directly on the tank data set, the current error message seems to be caused by the size of the input image.
Looking forward to your reply
@YuhsiHu @wangchengze001 @xy-guo 请教一下，这个网络是不能测试tank and temples数据集嘛？为啥都是稀疏点
那个相机内参需要调整一下，我很久没做实验了，文件找不到了，就记着是这个解决思路了。
---Original--- From: @.> Date: Sun, Nov 6, 2022 21:07 PM To: @.>; Cc: @.@.>; Subject: Re: [xy-guo/MVSNet_pytorch] Terrible reconstruction effect onTankandTemples dataset (#23)
@YuhsiHu @wangchengze001 @xy-guo 请教一下，这个网络是不能测试tank and temples数据集嘛？为啥都是稀疏点
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>
感谢您的回复，我对eval.py文件做了如下修改： intrinsics[:2, :] /= 4 color = ref_img[1:-24:4, 1::4, :][valid_points]
改成了大于0.6 view=3 得到下边的效果，看起来并不是很好，请问该修改什么参数提高mvsnet_pytorch在tt数据集上的的效果
------------------ 原始邮件 ------------------ 发件人: "xy-guo/MVSNet_pytorch" @.>; 发送时间: 2022年11月8日(星期二) 下午2:01 @.>; @.@.>; 主题: Re: [xy-guo/MVSNet_pytorch] Terrible reconstruction effect on TankandTemples dataset (#23)
那个相机内参需要调整一下，我很久没做实验了，文件找不到了，就记着是这个解决思路了。
---Original--- From: @.> Date: Sun, Nov 6, 2022 21:07 PM To: @.>; Cc: @.@.>; Subject: Re: [xy-guo/MVSNet_pytorch] Terrible reconstruction effect onTankandTemples dataset (#23)
@YuhsiHu @wangchengze001 @xy-guo 请教一下，这个网络是不能测试tank and temples数据集嘛？为啥都是稀疏点
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.> — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.>
