Here we implement EfficientDet. The code is based on a RetinaNet implementation by yhenon/pytorch-retinanet. We use the EfficientNet backend by rwightman/gen-efficientnet-pytorch.
Current implementation is able to run. I'll update this document as soon as I have some preliminary results. The paper by Tan et al. gives a few more details, which we would like to implement and report on:
- Use exponential moving average with decay 0.9998.
- Initialize convolution layers
- Train model using using SGD optimizer with momentum 0.9 and weight decay 4e-5.
- Implement described learning rate, which is first linearly increased from 0 to 0.08 in the initial 5% warm-up training steps and then annealed down using cosine decay rule.
- Report performance.
If you have other issues that need my attention, feel free to make a pull request or leave an issue.
Model | mAP | #Params | #FLOPS
-
Clone this repo
-
Install the required packages:
apt-get install tk-dev python-tk
- Install the python packages:
pip install pandas
pip install pycocotools
pip install cython
pip install opencv-python
pip install requests
pip install efficientnet_pytorch
Note that you may have to edit line 14 of build.sh if you want to change which version of python you are building the extension for.
The network can be trained using the train.py script. Currently, two dataloaders are available: COCO and CSV. For training on coco, use
python3 train.py --efficientnet --dataset coco --coco_path ../../Datasets/COCO2017 --scaling-compound 0 --batch-size 8
For training using a custom dataset, with annotations in CSV format (see below), use
python train.py --dataset csv --csv_train <path/to/train_annots.csv> --csv_classes <path/to/train/class_list.csv> --csv_val <path/to/val_annots.csv>
Note that the --csv_val argument is optional, in which case no validation will be performed.
- The code is based on a RetinaNet implementation by yhenon/pytorch-retinanet.
- Significant amounts of code are borrowed from the keras retinanet implementation
- The NMS module used is from the pytorch faster-rcnn implementation
- We use the EfficientNet backend by rwightman/gen-efficientnet-pytorch.