Deng Cai, Liang Peng, Liang Xie, Chenxi Huang, Chengqi Deng, Haifeng Liu, Xiaopei Wu, Honghui Yang
We lifted 13 functions out of this paper's own repositories and ran 5 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| LittlePey/SFD | canonical | 5 of 13 |
| Function | Status | Where it lives |
|---|---|---|
| get_lidar_in_image_fov | Ran | LittlePey/SFD/kitti_object.py code served (permissive licence) · get_code("47397a2377a470c7") |
| random_flip_along_y | Ran | LittlePey/SFD/pcdet/datasets/augmentor/augmentor_utils_sfd.py code served (permissive licence) · get_code("aa16fbdd4a5afcf1") |
| rotx | Ran | LittlePey/SFD/kitti_util.py code served (permissive licence) · get_code("8843d80ab666c4e9") |
| roty | Ran | LittlePey/SFD/kitti_util.py code served (permissive licence) · get_code("ae86074e10e1dfd7") |
| rotz | Ran | LittlePey/SFD/kitti_util.py code served (permissive licence) · get_code("082920adf6e29c94") |
| cfg_from_yaml_file | Not yet run | LittlePey/SFD/pcdet/config.py code served (permissive licence) · get_code("696fe155f9990d38") |
| depth_read | Not yet run | LittlePey/SFD/depth_to_lidar.py code served (permissive licence) · get_code("40c0b7bcfe33fd6d") |
| get_corner_loss_lidar | Not yet run | LittlePey/SFD/pcdet/utils/loss_utils.py code served (permissive licence) · get_code("1780d388cc532a6d") |
| get_global_grid_points_of_roi | Not yet run | LittlePey/SFD/pcdet/utils/loss_utils.py code served (permissive licence) · get_code("269fcc858076aa5f") |
| get_gridify_iou3d_loss | Not yet run | LittlePey/SFD/pcdet/utils/loss_utils.py code served (permissive licence) · get_code("f133fe553902b4ff") |
| global_rotation | Not yet run | LittlePey/SFD/pcdet/datasets/augmentor/augmentor_utils_sfd.py code served (permissive licence) · get_code("79d4891f6b00a1cd") |
| merge_new_config | Not yet run | LittlePey/SFD/pcdet/config.py code served (permissive licence) · get_code("392c0cf3a1b07b12") |
| random_flip_along_x | Not yet run | LittlePey/SFD/pcdet/datasets/augmentor/augmentor_utils_sfd.py code served (permissive licence) · get_code("00692f8182ba51ca") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Current LiDAR-only 3D detection methods inevitably suffer from the sparsity of point clouds. Many multi-modal methods are proposed to alleviate this issue, while different representations of images and point clouds make it difficult to fuse them, resulting in suboptimal performance. In this paper, we present a novel multi-modal framework SFD (Sparse Fuse Dense), which utilizes pseudo point clouds generated from depth completion to tackle the issues mentioned above. Different from prior works, we propose a new RoI fusion strategy 3D-GAF (3D Grid-wise Attentive Fusion) to make fuller use of information from different types of point clouds. Specifically, 3D-GAF fuses 3D RoI features from the pair of point clouds in a grid-wise attentive way, which is more fine-grained and more precise. In addition, we propose a SynAugment (Synchronized Augmentation) to enable our multi-modal framework to utilize all data augmentation approaches tailored to LiDAR-only methods. Lastly, we customize an effective and efficient feature extractor CPConv (Color Point Convolution) for pseudo point clouds. It can explore 2D image features and 3D geometric features of pseudo point clouds simultaneously. Our method holds the highest entry on the KITTI car 3D object detection leaderboard † , demonstrating the effectiveness of our SFD. Code will be made publicly available.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2203.09780")
get_code_for_paper("2203.09780")
have("2203.09780")
Connect an agent — have() is free.