GFL (Generalized Focal Loss) is an object detection model that utilizes an improved focal loss function to address the class imbalance problem, enhancing classification capability and improving the detection accuracy of multi-scale objects and the precision of bounding box predictions. It is suitable for object detection tasks in complex scenes.
| GPU | IXUCA SDK | Release | Branch |
|---|---|---|---|
| MR-V100 | 4.4.0 | 26.03 | release/26.03 |
| MR-V100 | 4.3.0 | 25.12 | release/25.12 |
Note: 请切换到与您的 SDK 版本对应的 Release 分支进行测试。请勿直接在 master 分支上运行测试，因为 master 分支可能包含与您的本地 SDK 版本不兼容的最新更改。
切换分支命令示例：
git checkout release/26.03
Pretrained model: https://download.openmmlab.com/mmdetection/v2.0/gfl/gfl_r50_fpn_1x_coco/gfl_r50_fpn_1x_coco_20200629_121244-25944287.pth
Dataset:
- https://github.com/ultralytics/assets/releases/download/v0.0.0/coco2017labels.zip to download the labels dataset.
- http://images.cocodataset.org/zips/val2017.zip to download the validation dataset.
- http://images.cocodataset.org/zips/train2017.zip to download the train dataset.
unzip -q -d ./ coco2017labels.zip
unzip -q -d ./coco/images/ train2017.zip
unzip -q -d ./coco/images/ val2017.zip
coco
├── annotations
│ └── instances_val2017.json
├── images
│ ├── train2017
│ └── val2017
├── labels
│ ├── train2017
│ └── val2017
├── LICENSE
├── README.txt
├── test-dev2017.txt
├── train2017.cache
├── train2017.txt
├── val2017.cache
└── val2017.txtContact the Iluvatar administrator to get the missing packages:
- mmcv-*.whl
# Install libGL
## CentOS
yum install -y mesa-libGL
## Ubuntu
apt install -y libgl1-mesa-glx
pip3 install -r requirements.txt# export onnx model
python3 export.py --weight gfl_r50_fpn_1x_coco_20200629_121244-25944287.pth --cfg gfl_r50_fpn_1x_coco.py --output gfl.onnx
# use onnxsim optimize onnx model
onnxsim gfl.onnx gfl_opt.onnxexport DATASETS_DIR=/Path/to/coco/# Accuracy
bash scripts/infer_gfl_fp16_accuracy.sh
# Performance
bash scripts/infer_gfl_fp16_performance.sh| Model | BatchSize | Precision | FPS | IOU@0.5 | IOU@0.5:0.95 |
|---|---|---|---|---|---|
| GFL | 32 | FP16 | 139.78 | 0.552 | 0.378 |