Transformer-Series

Introduction

This project is mainly used to complete the paper reproduction and precision alignment, it will be updated continuously. At present, it mainly supports the DETR series of object detection algorithms.The code continues the simple style of Facebook AI Research, which is mainly for the convenience of subsequent maintenance and learning.

Environment

• CUDA > 11.1 torch >= 1.10.1 torchvision >= 0.10.1
• numpy = 1.23.5 pycocotools = 2.0 scipy = 1.7.1

What's New

• [23.4.9] Support DINO-Scale4 .
• [23.4.4] Support Visualize Tools and DN-Deformable-DETR .
• [23.4.1] Support DN-DETR.
• Support DAB-DETR and DAB-Deformable-DETR .

Model List

Detection methods

• ✅ DETR
• ✅ Deformable-DETR
• ✅ Conditional-DETR
• ✅ DAB-DETR
• ✅ DAB-Deformable-DETR
• ✅ DN-DETR
• ✅ DINO

Segmentation methods

• [MaskDINO](comming soon)
• [MaskFormer]
• [Mask2Former]

Usage - Object detection

First, clone the repository locally:

git clone 

Then, install PyTorch 1.5+ and torchvision 0.6+:

conda install -c pytorch pytorch torchvision

Install pycocotools (for evaluation on COCO) and scipy (for training):

conda install cython scipy
pip install -U 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'

Install other requirements:

pip install -r requirements.txt

That's it, should be good to train and evaluate detection models.

(optional) to work with panoptic install panopticapi:

pip install git+https://github.com/cocodataset/panopticapi.git

Compiling CUDA operators

cd ./model/basic_operator
python setup.py build install
# unit test (should see all checking is True)
python test.py

Usage - Visualization

Used to visualize feature maps as well as attention map
See test.py for details

Sample:

Usage - Segmentation

Comming Soon !!

Data preparation

Download and extract COCO 2017 train and val images with annotations from http://cocodataset.org. We expect the directory structure to be the following:

path/to/coco/
  annotations/  # annotation json files
  train2017/    # train images
  val2017/      # val images

Training

To train baseline DETR on a single node with 8 gpus for 300 epochs run:

python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py --coco_path /path/to/coco 

Evaluation

To evaluate DETR R50 on COCO val5k with a single GPU run:

python main.py --batch_size 2 --no_aux_loss --eval --resume /path/to/checkpoints --coco_path /path/to/coco