@@ -60,30 +60,30 @@ We follow the below style to name config files. Contributors are advised to foll
60
60
In MMDetection3D's root directory, run the following command to train the model:
61
61
62
62
``` bash
63
- python tools/train.py projects/CenterFormer/configs/centerformer_voxel01_second-atten_secfpn-atten_4xb4-cyclic-20e_waymoD5-3d-class .py
63
+ python tools/train.py projects/CenterFormer/configs/centerformer_voxel01_second-atten_secfpn-atten_4xb4-cyclic-20e_waymoD5-3d-3class .py
64
64
```
65
65
66
66
For multi-gpu training, run:
67
67
68
68
``` bash
69
- python -m torch.distributed.launch --nnodes=1 --node_rank=0 --nproc_per_node=${NUM_GPUS} --master_port=29506 --master_addr=" 127.0.0.1" tools/train.py projects/CenterFormer/configs/centerformer_voxel01_second-atten_secfpn-atten_4xb4-cyclic-20e_waymoD5-3d-class .py
69
+ python -m torch.distributed.launch --nnodes=1 --node_rank=0 --nproc_per_node=${NUM_GPUS} --master_port=29506 --master_addr=" 127.0.0.1" tools/train.py projects/CenterFormer/configs/centerformer_voxel01_second-atten_secfpn-atten_4xb4-cyclic-20e_waymoD5-3d-3class .py
70
70
```
71
71
72
72
### Testing commands
73
73
74
74
In MMDetection3D's root directory, run the following command to test the model:
75
75
76
76
``` bash
77
- python tools/train.py projects/CenterFormer/configs/centerformer_voxel01_second-atten_secfpn-atten_4xb4-cyclic-20e_waymoD5-3d-class .py ${CHECKPOINT_PATH}
77
+ python tools/train.py projects/CenterFormer/configs/centerformer_voxel01_second-atten_secfpn-atten_4xb4-cyclic-20e_waymoD5-3d-3class .py ${CHECKPOINT_PATH}
78
78
```
79
79
80
80
## Results and models
81
81
82
82
### Waymo
83
83
84
- | Backbone | Load Interval | Voxel type (voxel size) | Multi-Class NMS | Multi-frames | Mem (GB) | Inf time (fps) | mAP@L1 | mAPH@L1 | mAP@L2 | ** mAPH@L2** | Download |
85
- | :----------------------------------------------------------------------------------------------------------------: | :-----------: | :---------------------: | :-------------: | :----------: | :------: | :------------: | :----: | :-----: | :----: | :---------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
86
- | [ SECFPN_WithAttention] ( ./configs/centerformer_voxel01_second-attn_secfpn-attn_4xb4-cyclic-20e_waymoD5-3d-class .py ) | 5 | voxel (0.1) | ✓ | × | 14.8 | | 72.2 | 69.5 | 65.9 | 63.3 | [ log] ( https://download.openmmlab.com/mmdetection3d/v1.1.0_models/centerformer/centerformer_voxel01_second-attn_secfpn-attn_4xb4-cyclic-20e_waymoD5-3d-class /centerformer_voxel01_second-attn_secfpn-attn_4xb4-cyclic-20e_waymoD5-3d-class_20221227_205613 -70c9ad37.json ) |
84
+ | Backbone | Load Interval | Voxel type (voxel size) | Multi-Class NMS | Multi-frames | Mem (GB) | Inf time (fps) | mAP@L1 | mAPH@L1 | mAP@L2 | ** mAPH@L2** | Download |
85
+ | :----------------------------------------------------------------------------------------------------------------- : | :-----------: | :---------------------: | :-------------: | :----------: | :------: | :------------: | :----: | :-----: | :----: | :---------: | :- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
86
+ | [ SECFPN_WithAttention] ( ./configs/centerformer_voxel01_second-attn_secfpn-attn_4xb4-cyclic-20e_waymoD5-3d-3class .py ) | 5 | voxel (0.1) | ✓ | × | 14.8 | | 72.2 | 69.5 | 65.9 | 63.3 | [ log] ( https://download.openmmlab.com/mmdetection3d/v1.1.0_models/centerformer/centerformer_voxel01_second-attn_secfpn-attn_4xb4-cyclic-20e_waymoD5-3d-3class /centerformer_voxel01_second-attn_secfpn-attn_4xb4-cyclic-20e_waymoD5-3d-3class_20221227_205613 -70c9ad37.log ) |
87
87
88
88
** Note** that ` SECFPN_WithAttention ` denotes both SECOND and SECONDFPN with ChannelAttention and SpatialAttention.
89
89
0 commit comments