PV-RCNN代码测试——gt_annos & dt_annos

PV-RCNN：paper，code

打印gt_annos和dt_annos代码

# pcdet/datasets/kitti/kitti_object_eval_python/eval.py
print('============================ details of evaluation of image6 start====================')
print('gt_annos:',gt_annos)
print('dt_annos:',dt_annos)
print('overlaps:',overlaps)
print('============================ details of evaluation of image6 end====================')

采用的测试集

# data/kitti/ImageSets/val.txt
000006

开始测试

python test.py --cfg_file cfgs/kitti_models/pv_rcnn.yaml --batch_size 1 --ckpt ../output/kitti_models/pv_rcnn/mydata_1/ckpt/checkpoint_epoch_10.pth --extra_tag 'mydata_2'

输出结果

============================ details of evaluation of image6 start====================
gt_annos: 
[{
'name': array(['Car', 'Car', 'Car', 'Car', 'DontCare', 'DontCare'], dtype='<U8'), 
'truncated': array([ 0.,  0.,  0.,  0., -1., -1.]), 
'occluded': array([ 2.,  0.,  0.,  1., -1., -1.]), 
'alpha': array([ -1.55,  -1.21,   0.15,   2.05, -10.  , -10.  ]), 
'bbox': 
array([[548.  , 171.33, 572.4 , 194.42],
       [505.25, 168.37, 575.44, 209.18],
       [ 49.7 , 185.65, 227.42, 246.96],
       [328.67, 170.65, 397.24, 204.16],
       [603.36, 169.62, 631.06, 186.56],
       [578.97, 168.88, 603.78, 187.56]], dtype=float32), 
'dimensions': 
array([[ 3.62,  1.48,  1.56],
       [ 4.32,  1.67,  1.64],
       [ 3.88,  1.5 ,  1.62],
       [ 4.29,  1.68,  1.67],
       [-1.  , -1.  , -1.  ],
       [-1.  , -1.  , -1.  ]]), 
'location': 
array([[-2.720e+00,  8.200e-01,  4.822e+01],
       [-2.610e+00,  1.130e+00,  3.173e+01],
       [-1.254e+01,  1.640e+00,  1.972e+01],
       [-1.266e+01,  1.130e+00,  3.844e+01],
       [-1.000e+03, -1.000e+03, -1.000e+03],
       [-1.000e+03, -1.000e+03, -1.000e+03]], dtype=float32), 
'rotation_y': array([ -1.62,  -1.3 ,  -0.42,   1.73, -10.  , -10.  ]), 
'score': array([-1., -1., -1., -1., -1., -1.]), 
'difficulty': array([-1,  0,  0,  1, -1, -1], dtype=int32), 
'index': array([ 0,  1,  2,  3, -1, -1], dtype=int32), 
'gt_boxes_lidar': 
array([[48.47077179,  2.95906162, -0.25629205,  3.62      ,  1.56      , 1.48      ,  0.04920367],
       [31.9806385 ,  2.77186871, -0.42351833,  4.32      ,  1.64      , 1.67      , -0.27079633],
       [19.92111015, 12.64801311, -0.87546992,  3.88      ,  1.62      , 1.5       , -1.15079633],
       [38.64176178, 12.85393906, -0.3297669 ,  4.29      ,  1.67      , 1.68      , -3.30079633]]), 
'num_points_in_gt': array([  9,  64, 321,  26,  -1,  -1], dtype=int32)
}]

dt_annos: 
[{
'name': array(['Car', 'Car', 'Car', 'Car', 'Car', 'Car', 'Car', 'Pedestrian', 'Car', 'Cyclist', 'Car'], dtype='<U10'), 
'truncated': array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.]), 
'occluded': array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.]), 
'alpha': 
array([-6.103507 , -7.479403 , -7.229924 , -7.605979 , -7.675108 ,
       -7.8643575, -6.357685 , -3.6272836, -7.6954   , -7.394093 ,
       -7.70753  ], dtype=float32), 
'bbox': 
array([[ 35.64774 , 183.85912 , 223.3187  , 243.51346 ],
       [507.01978 , 169.10129 , 575.2891  , 208.25652 ],
       [327.06696 , 179.01222 , 396.99054 , 207.22461 ],
       [579.0248  , 168.23694 , 606.95654 , 186.32397 ],
       [604.25714 , 168.71461 , 630.9089  , 188.09676 ],
       [554.41064 , 170.54095 , 578.0972  , 192.97998 ],
       [  0.      , 169.64636 , 133.15398 , 211.0606  ],
       [767.56995 , 167.53271 , 775.1517  , 193.86832 ],
       [621.69293 , 162.00407 , 646.0002  , 181.39088 ],
       [  0.      , 173.6811  ,  26.973091, 215.62674 ],
       [  0.      , 177.35478 ,  40.594414, 208.54866 ]], dtype=float32), 
'dimensions': 
array([[4.051301  , 1.4906676 , 1.6367772 ],
       [4.1166224 , 1.6087878 , 1.6226968 ],
       [3.876073  , 1.4151212 , 1.5551785 ],
       [3.9821217 , 1.5950358 , 1.6364912 ],
       [3.8610003 , 1.5991639 , 1.632775  ],
       [3.5636995 , 1.4554182 , 1.5348462 ],
       [4.695579  , 1.728069  , 1.7498904 ],
       [0.35432252, 1.9271181 , 0.4870359 ],
       [4.127525  , 1.8036367 , 1.724487  ],
       [1.8084677 , 1.7363664 , 0.4653614 ],
       [3.7997792 , 1.626316  , 1.6570537 ]], dtype=float32), 
'location': 
array([[-1.2713813e+01,  1.5605631e+00,  1.9605337e+01],
       [-2.5947142e+00,  1.0996835e+00,  3.1718645e+01],
       [-1.2629819e+01,  1.2905513e+00,  3.8112820e+01],
       [-7.0229030e-01,  4.2516139e-01,  6.5473595e+01],
       [ 1.4379555e+00,  5.4409987e-01,  6.1320782e+01],
       [-2.3913438e+00,  7.4463969e-01,  4.8419548e+01],
       [-2.3906298e+01,  1.2337779e+00,  3.2173038e+01],
       [ 1.2514790e+01,  9.0499169e-01,  5.2845219e+01],
       [ 3.1855197e+00, -1.1204358e-02,  6.9050949e+01],
       [-2.5308802e+01,  1.4130213e+00,  3.0669430e+01],
       [-3.2187435e+01,  1.4103941e+00,  3.9506737e+01]], dtype=float32), 
'rotation_y': 
array([-6.677984 , -7.565404 , -7.552907 , -7.6215177, -7.6566405, 
       -7.918352 , -6.9978104, -3.4008198, -7.65427  , -8.084847 , 
       -8.392922 ], dtype=float32), 
'score': 
array([0.9947767 , 0.9884449 , 0.87549245, 0.7574657 , 0.7471658 ,
       0.6317981 , 0.5018368 , 0.33131832, 0.170293  , 0.12664953,
       0.11167401], dtype=float32), 
'boxes_lidar': 
array([[ 19.805847  ,  12.820396  ,  -1.5438182 ,   4.051301  ,  1.6367772 ,   1.4906676 ,   5.1071877 ],
       [ 31.96945   ,   2.7562013 ,  -1.2283354 ,   4.1166224 ,  1.6226968 ,   1.6087878 ,   5.9946074 ],
       [ 38.314247  ,  12.823895  ,  -1.3296634 ,   3.876073  ,  1.5551785 ,   1.4151212 ,   5.9821105 ],
       [ 65.73507   ,   1.0215081 ,  -0.6745052 ,   3.9821217 ,  1.6364912 ,   1.5950358 ,   6.050721  ],
       [ 61.59235   ,  -1.1375881 ,  -0.8042521 ,   3.8610003 ,  1.632775  ,   1.5991639 ,   6.085844  ],
       [ 48.67214   ,   2.6305897 ,  -0.92508245,   3.5636995 ,  1.5348462 ,   1.4554182 ,   6.3475556 ],
       [ 32.32002   ,  24.069956  ,  -1.133335  ,   4.695579  ,  1.7498904 ,   1.728069  ,   5.427014  ],
       [ 53.169582  , -12.251147  ,  -1.2597351 ,   0.35432252,  0.4870359 ,   1.9271181 ,   1.8300234 ],
       [ 69.33254   ,  -2.8532667 ,  -0.29082227,   4.127525  ,  1.724487  ,   1.8036367 ,   6.0834737 ],
       [ 30.80909   ,  25.466953  ,  -1.2929395 ,   1.8084677 ,  0.4653614 ,   1.7363664 ,   6.514051  ],
       [ 39.612885  ,  32.38824   ,  -1.242166  ,   3.7997792 ,  1.6570537 ,   1.626316  ,   6.8221264 ]], dtype=float32), 
'frame_id': '000006'
}]

overlaps: 
[
array([[0.        , 0.        , 0.78456026, 0.        , 0.        , 0.        ],
       [0.        , 0.89253873, 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.50814879, 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.54982352, 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ],
       [0.        , 0.        , 0.        , 0.        , 0.        , 0.        ]])
]
============================ details of evaluation of image6 end====================

从overlaps中可以看出gt和dt的对应关系为

pic: 6 , total_dt_num: 11 , total_gt_num: 6
dt idx: 0 , gt idx: 2 , iou: 0.7845602631568909
dt idx: 1 , gt idx: 1 , iou: 0.8925387263298035
dt idx: 2 , gt idx: 3 , iou: 0.5081487894058228
dt idx: 5 , gt idx: 0 , iou: 0.549823522567749

以0号dt和2号gt为例，具体数据为

dt_annos: 
[{
'name': 'Car', 
'truncated':0., 
'occluded': 0., 
'bbox': [ 35.64774 , 183.85912 , 223.3187  , 243.51346 ], 
'dimensions': [4.051301  , 1.4906676 , 1.6367772 ], 
'location': [-1.2713813e+01,  1.5605631e+00,  1.9605337e+01], 
'score':0.9947767 , 
}]

gt_annos: 
[{
'name': 'Car', 
'truncated': 0., 
'occluded':  0., 
'bbox':[ 49.7 , 185.65, 227.42, 246.96],
'dimensions': [ 3.88,  1.5 ,  1.62],
'location': [-1.254e+01,  1.640e+00,  1.972e+01],
'score': -1., 
'difficulty':  0, 
'index': 2, 
}]

我的其他PV-RCNN代码测试系列文章，如果对你有帮助的话，请给我点赞哦~
PV-RCNN代码测试——TP，FP，TN，FN的计算（一）
PV-RCNN代码测试——TP，FP，TN，FN的计算（二）
PV-RCNN代码测试——gt_annos & dt_annos
PV-RCNN代码测试——计算iou