Skip to content

Latest commit

 

History

History
66 lines (50 loc) · 2.95 KB

readme.md

File metadata and controls

66 lines (50 loc) · 2.95 KB
Raw

mmdet使用教程

mmdet教程命令

  1. conda create -n mmdet_py39 python=3.9 anaconda
  2. https://mmdetection.readthedocs.io/en/latest/get_started.html
  3. https://pytorch.org/get-started/previous-versions/
    pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 --index-url https://download.pytorch.org/whl/cu121
  4. https://mmdetection.readthedocs.io/zh-cn/latest/user_guides/train.html#id7

mmdet运行命令

  1. 训练

     python tools/train.py <your-config-file>

  2. 测试

     python tools/test.py <your-config-file> <your-model-weights-file> --out <save-pickle-path>

  3. 计算量、参数量计算脚本

     python tools/analysis_tools/get_flops.py <your-config-file>

  4. 推理时间、fps、gpu memory计算脚本

     python tools/analysis_tools/benchmark.py <your-config-file> --checkpoint <your-model-weights-file> --task inference --fuse-conv-bn

  5. 绘制曲线图脚本

     python tools/analysis_tools/analyze_logs.py plot_curve <train-json-file> --keys <keys> --legend <legend> --out <save-path>

  6. 结果分析脚本

     python tools/analysis_tools/analyze_results.py <your-config-file> <test-pickle-path> <save-path>

mmdet视频教程链接(可按顺序观看)

  1. 一库打尽目标检测对比实验！mmdetection环境、训练、测试手把手教程！
  2. 一库打尽目标检测对比实验！mmdetection参数量、计算量、FPS、绘制logs手把手教程
  3. 一库打尽目标检测对比实验！mmdetection指标转换YOLO指标！

mmdet实验数据(指标均为COCO指标)

以下实验数据环境:
python:3.9.19
torch:2.1.0+cu121
torchvision:0.16.0
mmdet:3.3.0
mmcv:2.1.0
mmengine:0.10.3
硬件环境:
Platform:Ubuntu
CPU:i7-12700K
RAM:32G
GPU:RTX3090

VisDrone2019-testset

model Input Shape GFlops Params coco/bbox_mAP coco/bbox_mAP_50 coco/bbox_mAP_s coco/bbox_mAP_m coco/bbox_mAP_l
Faster-RCNN-R50-FPN-CIOU (768, 1344) 208G 41.39M 0.194 0.329 0.095 0.309 0.429
Cascade-RCNN-R50-FPN (768, 1344) 236G 69.29M 0.197 0.326 0.099 0.309 0.406
ATSS-R50-FPN-DyHead (768, 1344) 110G 38.91M 0.204 0.338 0.100 0.317 0.485
TOOD-R50 (768, 1344) 199G 32.04M 0.204 0.339 0.102 0.317 0.403
DINO (750, 1333) 274G 47.56M 0.253 0.445 0.150 0.371 0.503
DDQ (768, 1333) - - 0.268 0.463 0.159 0.390 0.526
YOLOX-Tiny (640, 640) 7.578G 5.035M 0.148 0.278 0.076 0.221 0.278
GFL (768, 1344) 206G 32.279M 0.193 0.321 0.094 0.300 0.409
RTMDet-Tiny (640, 640) 8.033G 4.876M 0.184 0.312 0.077 0.288 0.445
RetinaNet-R50-FPN (768, 1344) 210G 36.517M 0.164 0.276 0.060 0.274 0.427