Matlab code

[BiaBibii]

Hello, how do I need to run the matlab codes in order to obtain the accuracy and the overall?

[wangchengze001]

hello，i want request a question for matlab code ， are you reuse it to get accuracy？

[BiaBibii]

Hello, yes. I used this code to get accuracy and the overall. https://github.com/xy-guo/MVSNet_pytorch/tree/master/evaluations/dtu

[wangchengze001]

please give some advices in code，because i see 7 .m document in dtu ，i‘m not good at matlab ， just download matlab and activate，could you tell me how to change code in order to run it to get the acc & overall？i just know the file path must be modify by correctly position , and i will appriciate for your help, this is very important to me and my gradute paper.

[BiaBibii]

You need to run BaseEvalMain_web.m and after ComputeStat_web.m in matlab terminal. When you download Matlab pay attention to also check there to download statistisc, otherwise you will get some errors.
You need to change the paths according to yours.

[wangchengze001]

thank you for your advices, very thanks
first, i install statistics to solve the machine learining errors
second, i change the path in BaseEvalMain_web.m & ComputeStat_web.m, detail: dataPath =(from mvs_training download ground truth pointcloud) , plyPath = (my model run eval.sh output pointcloud),and resultsPath = ( new named file to storing compute result ),but i encounter an error, shown below。

cSet =
1
DataInName =
'D:\master-experiment\0410_d128_epoch1_ply/mvsnet001_l3.ply'
EvalName =
'D:\master-experiment\evaluation_indicatorsmvsnet_Eval_1.mat'
D:\master-experiment\0410_d128_epoch1_ply/mvsnet001_l3.ply
ans =
11 38
历时 162.898124 秒。
downsample factor: 1.4051
历时 134.630598 秒。
错误使用 plyread
No such file or directory
出错 PointCompareMain (第 12 行)
StlMesh = plyread(StlInName); %STL points already reduced 0.2 mm neighbourhood density
出错 BaseEvalMain_web (第 47 行)
BaseEval=PointCompareMain(cSet,Qdata,dst,dataPath);

error code shown,
StlInName=[dataPath '/Points/stl/stl' sprintf('%03d',cSet) '_total.ply'];
#error line 12
StlMesh = plyread(StlInName); %STL points already reduced 0.2 mm neighbourhood density
Qstl=[StlMesh.vertex.x StlMesh.vertex.y StlMesh.vertex.z]';

i don‘t understand /Points/stl/stl means

Thank you for your guidance

[BiaBibii]

Download from here the Points: https://roboimagedata2.compute.dtu.dk/data/MVS/Points.zip (Points.zip\Points\stl)
https://roboimagedata2.compute.dtu.dk/data/MVS/SampleSet.zip (from this you'll get the OBSmask SampleSet.zip\SampleSet\MVS Data\ObsMask)
You need to download this folders mentioned, unzip and set the paths in the code.

[wangchengze001]

thank you, i already get acc and com in dtu evaluation, but it is not good as in mvsnet paper, i want to know if you evaluation tanks&temples dataset by mvsnet_pytorch? maybe i need to eval tank datasets.

[HiRenZeHui]

您好，我需要如何运行matlab代码才能获得准确性和整体性？

你请问您的问题解决了吗？能留个邮箱来具体讨论一下

[wangchengze001]

@HiRenZeHui 乐意与您交流 [email protected]

[xinchen-cx]

Hello, how do I need to run the matlab codes in order to obtain the accuracy and the overall?

Hi, sorry to bother you，when I use the offical dtu eval code, it report an error:
内存不足。
出错 KDTreeSearcher/rangesearch (line 149)
idx= internal.stats.KDTreeSearcher.knnsearchmex(X2, Y', [], minExp, obj.cutDim, obj.cutVal, ...
出错 reducePts_haa (line 22)
idx = rangesearch(NS,pts(:,RandOrd(Range))',dst);
出错 PointCompareMain (line 7)
Qdata=reducePts_haa(Qdata,dst);
出错 BaseEvalMain_web (line 60)
BaseEval=PointCompareMain(cSet,Qdata,dst,dataPath);

I can calculate the scene with few point clouds (. ply size of DTU scene 77 is 32M), but how can I calculate other scenes with hundreds M？