search

zhyever / SimIPU

[AAAI 2021] Official Implementation of "SimIPU: Simple 2D Image and 3D Point Cloud Unsupervised Pre-Training for Spatial-Aware Visual Representations"
Apache License 2.0
49 stars 6 forks source link

About intra-modal spatial perception module parameters update problem #11

Open jiangxb98 opened 1 year ago

jiangxb98 commented 1 year ago

Hello author, in the SimIPU paper, I find that the inter-modal module loss is gradient truncation in the point cloud features. When gradient backpropagation is considered, this step does not contribute to the update of the intra-modal module parameters. Why does adding this loss improve the point cloud 3D object detection performance?

jiangxb98 commented 1 year ago

Is the ablation experiment in Table 5 only used for monocular 3D object detection?