EfficientDet: Scalable and Efficient Object Detection, in PyTorch

A PyTorch implementation of EfficientDet from the 2019 paper by Mingxing Tan Ruoming Pang Quoc V. Le Google Research, Brain Team. The official and original: comming soon.

Fun with Demo:

python demo.py --weight ./checkpoint_VOC_efficientdet-d1_97.pth --threshold 0.6 --iou_threshold 0.5 --cam --score

Table of Contents

• Recent Update
• Benchmarking
• Installation
• Installation
• Prerequisites
• Datasets
• Train
• Evaluate
• Performance
• Demo
• Future Work
• Reference

Recent Update

• [06/01/2020] Support both DistributedDataParallel and DataParallel, change augmentation, eval_voc
• [17/12/2019] Add Fast normalized fusion, Augmentation with Ratio, Change RetinaHead, Fix Support EfficientDet-D0->D7
• [7/12/2019] Support EfficientDet-D0, EfficientDet-D1, EfficientDet-D2, EfficientDet-D3, EfficientDet-D4,... . Support change gradient accumulation steps, AdamW.

Benchmarking

We benchmark our code thoroughly on three datasets: pascal voc and coco, using family efficientnet different network architectures: EfficientDet-D0->7. Below are the results:

1). PASCAL VOC 2007 (Train/Test: 07trainval/07test, scale=600, ROI Align)

Installation

• Install PyTorch by selecting your environment on the website and running the appropriate command.
• Clone this repository and install package prerequisites below.
• Then download the dataset by following the instructions below.
• Note: For training, we currently support VOC and COCO, and aim to add ImageNet support soon.

prerequisites

• Python 3.6+
• PyTorch 1.3+
• Torchvision 0.4.0+ (We need high version because Torchvision support nms now.)
• requirements.txt

Datasets

To make things easy, we provide bash scripts to handle the dataset downloads and setup for you. We also provide simple dataset loaders that inherit torch.utils.data.Dataset, making them fully compatible with the torchvision.datasets API.

VOC Dataset

PASCAL VOC: Visual Object Classes

Download VOC2007 + VOC2012 trainval & test

# specify a directory for dataset to be downloaded into, else default is ~/data/
sh datasets/scripts/VOC2007.sh
sh datasets/scripts/VOC2012.sh

COCO

Microsoft COCO: Common Objects in Context

Download COCO 2017

# specify a directory for dataset to be downloaded into, else default is ~/data/
sh datasets/scripts/COCO2017.sh

Training EfficientDet

• To train EfficientDet using the train script simply specify the parameters listed in train.py as a flag or manually change them.

python train.py --network effcientdet-d0  # Example

• With VOC Dataset:

# DataParallel
python train.py --dataset VOC --dataset_root /root/data/VOCdevkit/ --network effcientdet-d0 --batch_size 32 
# DistributedDataParallel with backend nccl
python train.py --dataset VOC --dataset_root /root/data/VOCdevkit/ --network effcientdet-d0 --batch_size 32 --multiprocessing-distributed

• With COCO Dataset:

# DataParallel
python train.py --dataset COCO --dataset_root ~/data/coco/ --network effcientdet-d0 --batch_size 32
# DistributedDataParallel with backend nccl
python train.py --dataset COCO --dataset_root ~/data/coco/ --network effcientdet-d0 --batch_size 32 --multiprocessing-distributed

Evaluation

To evaluate a trained network:

• With VOC Dataset:

  python eval_voc.py --dataset_root ~/data/VOCdevkit --weight ./checkpoint_VOC_efficientdet-d0_261.pth

• With COCO Dataset comming soon.

Demo

python demo.py --threshold 0.5 --iou_threshold 0.5 --score --weight checkpoint_VOC_efficientdet-d1_34.pth --file_name demo.png

Output:

Webcam Demo

You can use a webcam in a real-time demo by running:

python demo.py --threshold 0.5 --iou_threshold 0.5 --cam --score --weight checkpoint_VOC_efficientdet-d1_34.pth

Performance

TODO

We have accumulated the following to-do list, which we hope to complete in the near future

• Still to come:
  • EfficientDet-[D0-7]
  • GPU-Parallel
  • NMS
  • Soft-NMS
  • Pretrained model
  • Demo
  • Model zoo
  • TorchScript
  • Mobile
  • C++ Onnx

Authors

• Toan Dao Minh

Note: Unfortunately, this is just a hobby of ours and not a full-time job, so we'll do our best to keep things up to date, but no guarantees. That being said, thanks to everyone for your continued help and feedback as it is really appreciated. We will try to address everything as soon as possible.

References

• tanmingxing, rpang, qvl, et al. "EfficientDet: Scalable and Efficient Object Detection." EfficientDet.
• A list of other great EfficientDet ports that were sources of inspiration:
  • EfficientNet
  • SSD.Pytorch
  • mmdetection
  • RetinaNet.Pytorch
  • NMS.Torchvision

Citation

@article{efficientdetpytoan,
    Author = {Toan Dao Minh},
    Title = {A Pytorch Implementation of EfficientDet Object Detection},
    Journal = {github.com/toandaominh1997/EfficientDet.Pytorch},
    Year = {2019}
}