打印信息
================ Start to Convert ================
================ Start to Copy Raw Data ================
================ Start to Generate Label ================
================ Start to Generate Calibration Files ================
15627 15627
================ Start to Generate ImageSet Files ================
您好，在DAIR-V2X-V数据集，通过下面的命令转为kitii格式： python dair2kitti.py --source-root ./data/DAIR-V2X-V/single-vehicle-side --target-root ./data/DAIR-V2X-V/single-vehicle-side --split-path ./data/split_datas/single-vehicle-split-data.json --label-type camera --sensor-view vehicle
打印信息 ================ Start to Convert ================ ================ Start to Copy Raw Data ================ ================ Start to Generate Label ================ ================ Start to Generate Calibration Files ================ 15627 15627 ================ Start to Generate ImageSet Files ================
查看calib中的相机标定文件，比如 000000.txt P2: 3996.487567 0.0 955.58618 0.0 0.0 3963.430994 527.646219 0.0 0.0 0.0 1.0 0.0 P2: 3996.487567 0.0 955.58618 0.0 0.0 3963.430994 527.646219 0.0 0.0 0.0 1.0 0.0 P2: 3996.487567 0.0 955.58618 0.0 0.0 3963.430994 527.646219 0.0 0.0 0.0 1.0 0.0 P2: 3996.487567 0.0 955.58618 0.0 0.0 3963.430994 527.646219 0.0 0.0 0.0 1.0 0.0 R0_rect: 1 0 0 0 1 0 0 0 1 Tr_velo_to_cam: 0.006283 -0.999979 -0.001899 -0.298036 -0.005334 0.001865 -0.999984 -0.666812 0.999966 0.006293 -0.005322 -0.516927 Tr_velo_to_cam: 0.006283 -0.999979 -0.001899 -0.298036 -0.005334 0.001865 -0.999984 -0.666812 0.999966 0.006293 -0.005322 -0.516927
查看lable_2中的标签，比如 000000.txt Car 0 0 0.33888581543844903 0 527.938232 69.723068 637.4556269999999 0.850836 4.337498 2.073565 -9.601712831407 0.8624079931420001 32.383280568744 1.615145
看起来转换应该梅问题的，但是，在可视化3D检测框时发现部分标签不准确，而且不准确的标签占比较大。 比如：000187.jpg
比如：000141.jpg
比如：000184.jpg
比如：000112.jpg
比如：000010.jpg
请问标签确实会存在偏差是吗