RTMDet-R: A robust instance segmentation network for complex traffic scenarios

This is an official implementations for "RTMDet-R: A robust instance segmentation network for complex traffic scenarios"

The main architecture of RTMDet-R is based on RTMDet. The code is based on mmdetection

Update

(4/14: Upload the code and pretrained model.)

Image results

Results comparision

Installation

Install the dependencies and prepare the datasets according to the guidelines in mmdetection

conda create -n rtmdetr python=3.8
conda activate rtmdetr
conda install pytorch torchvision -c pytorch
pip install -U openmim
mim install mmengine
mim install "mmcv>=2.0.0"
git clone https://github.com/open-mmlab/mmdetection.git
cd mmdetection
pip install -v -e .

Download our code using

git clone https://github.com/GTrui6/RTMDet-R.git

The BDD-C dataset is too large to upload. We here provide some of the data for visualize, which processed by TPSeNCE. We appreciate for their work in image generation, which helps us build the BDD-C dataset. The samples of BDD-C is here.

Visualization

You can require the weights of model here and do the model inference and visualization with

python demo/image_demo.py "path to your data image" \
        /configs/rtmdet-t/rtmdet-r_s.py "path to config file" \
        --palette voc \
        --weights "path to weights" \
        --out-dir "path to output"

Training

You can train your own dataset with the coco-style with

python tools/train.py    /configs/rtmdet-r/rtmdet-r_s.py \

Contact

For technical problem, please create an issue.

If you have any private question, please feel free to contact me via [email protected]