QRCode Detection

Deep learning based QRCode detection.

Introduction

This is a project which depends on deep learning algorithm for QRCode detection.
We have achieved fast and high-precision detection by using a yolov3-like detecter.

Feature:

Fast detection, more than 190 fps on GTX 1060.
High precision
Evaluate result on validation data

Precision Recall Mean IOU

0.987 0.819 0.798
Free deployment

Precision	Recall	Mean IOU
0.987	0.819	0.798

Installation

Please enable python in your machine.

git clone https://github.com/cosimo17/QRCodeDetection.git
cd QRCodeDetection
pip install -r requirements.txt

Test

To test with the pretrained model, please download the pretrained weight file from here.

python3 test.py \
	-w yolo_qrcode.h5 \
	-i test_images/1.jpg \
	-o ./result_1.jpg

Training

Before start training, please check How to prepare dataset
Run the kmean algorithm to generate priori anchor boxes

python3 utils/kmean.py \
		--root_dir your_dataset_dir \
		-n 6

Execute following command to start training:

python3 train.py \
	-d your_dataset_dir \
	-b 64 \
	-e 80

You can run python3 train.py --help to get help.
During training, you can use tensorboard to visualize the loss curve.

tensorboard --logdir=./logs

Evaluate

Execute following command to evaluate the model performance:

python3 evaluate.py \
	-d your_dataset_dir \
	-b 64 \
	--score_threshold 0.5 \
	--iou_threshold 0.5 \
	-w yolo_qrcode.h5