search

open-mmlab / OpenPCDet

OpenPCDet Toolbox for LiDAR-based 3D Object Detection.
Apache License 2.0
4.55k stars 1.28k forks source link

can someone tell me why my evaluation result is so high by pretrained model #1277

Closed RouDanCongJii closed 1 year ago

RouDanCongJii commented 1 year ago

i used pretrained second model and get the follow result 2023-02-23 18:59:26,709 INFO Car AP@0.70, 0.70, 0.70: bbox AP:99.3793, 90.7942, 90.4609 bev AP:99.2691, 90.4727, 89.9079 3d AP:98.8017, 89.7792, 89.0197 aos AP:99.37, 90.77, 90.42 Car AP_R40@0.70, 0.70, 0.70: bbox AP:99.7249, 97.1309, 92.0793 bev AP:99.6127, 94.3167, 91.5589 3d AP:99.2124, 93.4512, 88.3061 aos AP:99.72, 97.10, 92.04 Car AP@0.70, 0.50, 0.50: bbox AP:99.3793, 90.7942, 90.4609 bev AP:99.5096, 90.8294, 90.5235 3d AP:99.4960, 90.8083, 90.4818 aos AP:99.37, 90.77, 90.42 Car AP_R40@0.70, 0.50, 0.50: bbox AP:99.7249, 97.1309, 92.0793 bev AP:99.7907, 97.1785, 92.1358 3d AP:99.7853, 97.1409, 92.0966 aos AP:99.72, 97.10, 92.04 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:75.4804, 67.5479, 59.8178 bev AP:74.3144, 66.2717, 58.7973 3d AP:66.5938, 65.4820, 58.0839 aos AP:72.49, 64.58, 57.21 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:73.6872, 64.8776, 60.4409 bev AP:72.4613, 63.7811, 61.2739 3d AP:69.6311, 62.8962, 58.4841 aos AP:70.47, 61.75, 57.53 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:75.4804, 67.5479, 59.8178 bev AP:76.7871, 68.3064, 67.8505 3d AP:76.7725, 68.2853, 67.8399 aos AP:72.49, 64.58, 57.21 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:73.6872, 64.8776, 60.4409 bev AP:77.2685, 67.8897, 65.3601 3d AP:77.2587, 67.8793, 65.3485 aos AP:70.47, 61.75, 57.53 Cyclist AP@0.50, 0.50, 0.50: bbox AP:87.6958, 86.8851, 86.6909 bev AP:85.0618, 83.1781, 83.1824 3d AP:84.7369, 82.7035, 76.2567 aos AP:87.63, 86.68, 86.49 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.1682, 90.3902, 88.0840 bev AP:88.1020, 84.7045, 82.5290 3d AP:87.7432, 82.2763, 79.9919 aos AP:91.10, 90.15, 87.86 Cyclist AP@0.50, 0.25, 0.25: bbox AP:87.6958, 86.8851, 86.6909 bev AP:85.8410, 84.3228, 84.0087 3d AP:85.8410, 84.2733, 83.9586 aos AP:87.63, 86.68, 86.49 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.1682, 90.3902, 88.0840 bev AP:88.9074, 87.4259, 85.1907 3d AP:88.9074, 85.5976, 83.3479 aos AP:91.10, 90.15, 87.86

i think its so high. Should i modify any cfg file?

roger-lcc commented 1 year ago

I met the opposite situation. I use pre-trained pv_rcnn models and the low results. Car AP@0.70, 0.70, 0.70: bbox AP:95.6474, 89.2780, 88.7115 bev AP:89.8919, 87.3619, 86.0995 3d AP:88.9541, 78.9888, 78.1926 aos AP:95.63, 89.16, 88.53 Car AP_R40@0.70, 0.70, 0.70: bbox AP:97.3897, 93.7292, 91.6680 bev AP:92.6827, 88.3507, 87.7208 3d AP:91.1629, 82.4606, 79.9689 aos AP:97.37, 93.58, 91.46 Car AP@0.70, 0.50, 0.50: bbox AP:95.6474, 89.2780, 88.7115 bev AP:95.6423, 89.2949, 88.8683 3d AP:95.5877, 89.2724, 88.7918 aos AP:95.63, 89.16, 88.53 Car AP_R40@0.70, 0.50, 0.50: bbox AP:97.3897, 93.7292, 91.6680 bev AP:97.4131, 94.1802, 93.7167 3d AP:97.3637, 94.0607, 93.4874 aos AP:97.37, 93.58, 91.46 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:73.0045, 66.4030, 63.7966 bev AP:65.8625, 59.0644, 54.6418 3d AP:63.8111, 55.5421, 50.7981 aos AP:67.77, 61.25, 58.25 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:73.8617, 67.3097, 63.9010 bev AP:66.8020, 58.1495, 53.4317 3d AP:63.5816, 54.6249, 49.5536 aos AP:68.12, 61.40, 57.78 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:73.0045, 66.4030, 63.7966 bev AP:77.0930, 71.9638, 68.9911 3d AP:77.0287, 71.7998, 68.6653 aos AP:67.77, 61.25, 58.25 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:73.8617, 67.3097, 63.9010 bev AP:79.1872, 72.9949, 69.3870 3d AP:79.1127, 72.7706, 69.0667 aos AP:68.12, 61.40, 57.78 Cyclist AP@0.50, 0.50, 0.50: bbox AP:89.4385, 81.0023, 75.9043 bev AP:88.5275, 72.0754, 69.4044 3d AP:85.0159, 67.4382, 63.4889 aos AP:89.35, 80.63, 75.54 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:94.9273, 81.4149, 77.5261 bev AP:93.5859, 72.7094, 69.2519 3d AP:87.6054, 67.4265, 63.2141 aos AP:94.83, 81.03, 77.12 Cyclist AP@0.50, 0.25, 0.25: bbox AP:89.4385, 81.0023, 75.9043 bev AP:94.4824, 77.5773, 72.7021 3d AP:94.4824, 77.5712, 72.7021 aos AP:89.35, 80.63, 75.54 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:94.9273, 81.4149, 77.5261 bev AP:95.6758, 78.4157, 75.0008 3d AP:95.6758, 78.4140, 74.9767 aos AP:94.83, 81.03, 77.12 And I tried it in two PC(tecent cloud and local ubuntu). I got the close same results. And I thought it can the data. So I redownloaded the data from official website but I can't see the difference. Please help:)

github-actions[bot] commented 1 year ago

This issue is stale because it has been open for 30 days with no activity.

github-actions[bot] commented 1 year ago

This issue was closed because it has been inactive for 14 days since being marked as stale.