Skip to content
Codes for paper: "IoU-Uniform R-CNN: Breaking Through the Limitations of RPN"
Python Cuda C++ Other
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.github
configs
demo
docker
docs
mmdet
tests
tools add multi nodes distributed test support (#1399) Nov 29, 2019
.gitignore
.isort.cfg
.pre-commit-config.yaml
.style.yapf
.travis.yml
LICENSE
README.md
pytest.ini
requirements.txt
setup.py

README.md

IoU-Uniform R-CNN

This repository provides the codes for paper IoU-Uniform R-CNN

Setup

Please follow official installation and getting_started guides.

Training

./tools/dist_train.sh ${CONFIG_FILE} ${GPU_NUM} [--validate] [other_optional_args] Note:

  • Config files of IoU-Uniform R-CNN:

    • configs/faster_rcnn_r50_fpn_1x_IoU_reg.py
    • configs/faster_rcnn_r101_fpn_1x_IoU_reg.py
    • configs/cascade_rcnn_r50_fpn_IoU_reg_1x.py
    • configs/cascade_rcnn_r101_fpn_IoU_reg_1x.py
    • configs/pascal_voc/faster_rcnn_r50_fpn_1x_IoU_reg_voc0712.py
    • configs/pascal_voc/faster_rcnn_r50_fpn_1x_reg_separate_voc0712.py
    • configs/pascal_voc/cascade_rcnn_r50_fpn_1x_IoU_reg_voc0712.py
  • We train IoU-Uniform R-CNN and accompanied detectors with 2 GPUs and 2 img/GPU. According to the Linear Scaling Rule, you need to set the learning rate proportional to the batch size if you use different GPUs or images per GPU.

Testing

./tools/dist_test.sh ${CONFIG_FILE} ${CHECKPOINT_FILE} ${GPU_NUM} [--out ${RESULT_FILE}] [--eval ${EVAL_METRICS}]

TODO

  • Release IoU-Uniform R-CNN code base
  • Release trained models
You can’t perform that action at this time.