MMDetection on VOC2007 with Mask R-CNN & Sparse R-CNN

OpenMMLab MMDetection VOC2007 Mask R-CNN Sparse R-CNN

1.

Conda

bash conda create --name openmmlab python=3.8 -y conda activate openmmlab

PyTorchGPUCUDA 12.1

bash pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 --index-url https://download.pytorch.org/whl/cu121 python -c 'import torch;print(torch.__version__);print(torch.version.cuda)'

MMDetection

```bash pip install -U openmim mim install mmengine mim install "mmcv>=2.0.0"

git clone git@github.com:open-mmlab/mmdetection.git cd mmdetection pip install -v -e . ```

2.

VOC2007

bash wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar tar -xvf VOCtrainval_06-Nov-2007.tar tar -xvf VOCtest_06-Nov-2007.tar

COCO

bash python tools/dataset_converters/pascal_voc.py mmdetection/data/VOCdevkit --out-format coco -o mmdetection/data/coco

3.

• configs/_base_/models/mask-rcnn_r50_fpn.py num_classes=20
• configs/_base_/datasets/coco_instance.py
• configs/_base_/schedules/schedule_1x.py max_epochs=48
• mmdet/datasets/xml_style.py

4.

Mask R-CNN

bash export CUDA_VISIBLE_DEVICES=6 python tools/train.py mmdetection/configs/mask_rcnn/mask-rcnn_r50_fpn_1x_coco.py --work-dir work_dirs3/mask_rcnn_voc

Sparse R-CNN

bash export CUDA_VISIBLE_DEVICES=5,6,7 bash tools/dist_train.sh configs/sparse_rcnn/sparse-rcnn_r50_fpn_1x_coco.py 3 --work-dir work_dirs3/sparse_rcnn_voc

5. TensorBoard

```bash tensorboard --logdir mmdetection/workdirs3/maskrcnnvoc/20250526140843/vis_data --port 6007

tensorboard --logdir mmdetection/workdirs/sparsercnnvoc/20250527200710/vis_data --port 6008 ```

6. Proposal

bash python comparison/visualize_comparison.py --config configs/mask_rcnn/mask-rcnn_r50_fpn_1x_coco.py --checkpoint work_dirs3/mask_rcnn_voc/epoch_48.pth --output-dir comparison/proposal_results

7.

Mask R-CNN

bash python tools/test.py work_dirs3/mask_rcnn_voc/mask-rcnn_r50_fpn_1x_coco.py work_dirs3/mask_rcnn_voc/epoch_48.pth --show-dir comparison/mask/

Sparse R-CNN

bash python tools/test.py configs/sparse_rcnn/sparse-rcnn_r50_fpn_1x_coco.py work_dirs3/sparse_rcnn_voc/epoch_60.pth --show-dir comparison/sparse/

8.

Mask R-CNN

bash python demo/image_demo.py comparison/demo_image configs/mask_rcnn/mask-rcnn_r50_fpn_1x_coco.py --weights work_dirs/mask_rcnn_voc/epoch_48.pth --out-dir comparison/my_test/ --device cuda

Sparse R-CNN

bash python demo/image_demo.py comparison/demo_image configs/sparse_rcnn/sparse-rcnn_r50_fpn_1x_coco.py --weights work_dirs/sparse_rcnn_voc/epoch_60.pth --out-dir comparison/my_test_sparse/ --device cuda

mmdetection/ configs/ tools/ work_dirs3/ mask_rcnn_voc/ sparse_rcnn_voc/ comparison/ proposal_results/ mask/ sparse/ demo_image/ my_test_sparse/ my_test_mask/ data/ coco/

• MMDetection
• Pascal VOC
• OpenMMLab