PyTorch SSD

PyTorch implementation of SSD: Single Shot MultiBox Detector.

Results

PASCAL VOC

• Training: 07+12 trainval
• Evaluation: 07 test

Model	Input size	mAP0.5	Configuration
SSD300	300	77.1	configs/voc/ssd300.yaml
SSD512	512	79.4	configs/voc/ssd512.yaml
MobileNetV2 SSDLite	320	70.7	configs/voc/mobilenetV2_ssdlite.yaml

COCO

• Training: train2017
• Evaluation: val2017

Model	Input size	mAP0.5:0.95	Configuration
SSD300	300	25.3	configs/coco/ssd300.yaml
SSD512	512	29.4	configs/coco/ssd512.yaml

Note: We run coco_eval.py to obtain the COCO mAP scores, as described in Evaluation section. In coco_eval.py, pycocotools is used for mAP calculation.

Requirements

• Python ≥ 3.6
• Install libraries: pip install -r requirements.txt

Data Preparation

PASCAL VOC

cd datasets/voc/

wget http://pjreddie.com/media/files/VOCtrainval_06-Nov-2007.tar
wget http://pjreddie.com/media/files/VOCtest_06-Nov-2007.tar
wget http://pjreddie.com/media/files/VOCtrainval_11-May-2012.tar
tar xf VOCtrainval_06-Nov-2007.tar
tar xf VOCtest_06-Nov-2007.tar
tar xf VOCtrainval_11-May-2012.tar

python prepare.py --root VOCdevkit/

COCO

cd datasets/coco/

wget http://images.cocodataset.org/zips/train2017.zip
wget http://images.cocodataset.org/zips/val2017.zip
wget http://images.cocodataset.org/annotations/annotations_trainval2017.zip
unzip train2017.zip
unzip val2017.zip
unzip annotations_trainval2017.zip

python prepare.py --root .

Configuration

We use YAML for configuration management. See configs/*/*.yaml for examples. You can modify the settings as needed.

Training

python train.py --cfg <CONFIG_FILE> --logdir <LOG_DIRECTORY>

# For example, to train SSD300 on PASCAL VOC:
python train.py --cfg configs/voc/ssd300.yaml --logdir runs/voc_ssd300/exp0/

To visualize training progress using TensorBoard:

tensorboard --logdir <LOG_DIRECTORY>

An interrupted training can be resumed by:

# Run train.py with --resume to restore the latest saved checkpoint file in the log directory.
python train.py --cfg <CONFIG_FILE> --logdir <LOG_DIRECTORY> --resume

Evaluation

PASCAL VOC

python eval.py --cfg <CONFIG_FILE> --pth <LOG_DIRECTORY>/best.pth --dataset datasets/voc/val.json

COCO

Run coco_eval.py to calculate the COCO mAP metric using pycocotools:

python coco_eval.py --cfg <CONFIG_FILE> --pth <LOG_DIRECTORY>/best.pth --coco_dir <COCO_DIR>

Note: <COCO_DIR> can be datasets/coco/ if you prepare the COCO dataset according to the instructions in Data Preparation.