Why is the number of predicted point cloud labels not equal to the number of point clouds?

[xkhnhms]

Why is the number of predicted point cloud tags not equal to the number of point clouds?

---

./data/semantic_kitti/dataset/sequences_0.06/25/velodyne/1623230394502471.npy 1623230394502471.label (14316, 3) (26692,)

./data/semantic_kitti/dataset/sequences_0.06/25/velodyne/1623230394702421.npy 1623230394702421.label (14339, 3) (26729,)

[xkhnhms]

Hello @QingyongHu We use the main_SemanticKITTI.py you provided to test the dataset, and found that the number of predicted tags is inconsistent with the number of points in the point cloud data. Therefore, it is impossible to visualize to determine whether the test result is accurate. What do we need to do?

[QingyongHu]

Hi @xkhnhms, sorry for my late response,

Please check the code here, you need to back project the semantic labels to the raw point clouds and then calculate the accuracy and visualize.

Best, Qingyong