Error in ResultMerge.py

[Smoothing97]

按照readme走流程，特意用了example的几张图片，分割处理数据然后valid，得到了15个results文件，大概如下： P1234__1__0___2976 0.011820 696.137329 983.967957 756.505493 1024.000000 P1234__1__0___2976 0.010101 1.000000 924.940125 19.290693 969.360168 P1234__1__0___2976 0.008857 795.971069 960.030212 853.479736 1024.000000 P1234__1__0___2976 0.008307 1.000000 979.919922 124.899353 1024.000000 P1234__1__0___2976 0.010262 560.123779 828.110779 578.821289 863.667175 P1234__1__0___2976 0.010617 538.470886 988.313965 599.414612 1024.000000 P1234__1__0___2976 0.010275 764.724182 968.561584 819.250671 1024.000000 P1234__1__0___2976 0.009946 1004.529358 704.344116 1024.000000 733.913330

放到resultmerge.py，src_path=15个results的存放路径，dct_path=新建存放路径，总报错： (mytensorflow) aa@aa-GPU:~/DOTA_YOLOv2/DOTA_devkit$ python ResultMerge.py Traceback (most recent call last): File "ResultMerge.py", line 161, in mergebypoly(r'path_to_configure', r'path_to_configure') File "ResultMerge.py", line 158, in mergebypoly py_cpu_nms_poly) File "ResultMerge.py", line 127, in mergebase nameboxnmsdict = nmsbynamedict(nameboxdict, nms, nms_thresh) File "ResultMerge.py", line 76, in nmsbynamedict keep = nms(np.array(nameboxdict[imgname]), thresh) File "ResultMerge.py", line 12, in py_cpu_nms_poly scores = dets[:, 8] IndexError: index 8 is out of bounds for axis 1 with size 5

[dingjiansw101]

因为你的结果是Task2的，所以你应该使用mergebyrec()而不是mergebypoly

[Smoothing97]

谢谢您的指点，跑出结果了，但是用dota_evaluation_task2.py测map结果不太好： ... classname: ship check fp: [ 1. 1. 1. ..., 1. 1. 1.] check tp [ 0. 0. 0. ..., 0. 0. 0.] npos num: 555 ap: 1.37685416361e-05 ... classname: helicopter check fp: [ 1. 1. 1. ..., 1. 1. 1.] check tp [ 0. 0. 0. ..., 0. 0. 0.] npos num: 0 ap: 0.0 map: 9.21994456949e-06 classaps: [ 0.00310588 0.00575313 0. 0. 0. 0.00032483 0.00137685 0.00151971 0. 0.00174951 0. 0. 0. 0. 0. ] 困惑，我是按照教程摸索来的，但是不知道哪里出了问题

[dingjiansw101]

你是用的我们提供的模型跑的结果，还是自己训练的？

[Yanhong-Li]

我想问一下怎么获取15个results？