wuxiaolianggit/ctw-pytorch

Simplified Chinese text recognition implementation with yolov3 and pytorch

Recognize Chinese Characters on Traffic Signs

YOLOv3

Requirements

Python 3.7 or later with the following pip3 install -U -r requirements.txt packages:

• numpy
• torch >= 1.0.0
• opencv-python

Training

Gtx 1080 ti 12G RAM * 1

Start Training: Run train.py to begin training

Transfer Learning Run train.py --resume to start from pretrained weight

Multi-Scale Run train.py --multi-scale

num of class

1. traffic sign detection : 3 classes
2. Chinese text : 1 class

Image Augmentation Detail

Reference

Aug.	Description
Translation	+/- 10% (vertical and horizontal)
Rotation	+/- 5 degrees
Shear	+/- 2 degrees (vertical and horizontal)
Scale	+/- 10%
Reflection	50% probability (horizontal-only)
HSV Saturation	+/- 50%
HSV Intensity	+/- 50%
Distortation	+/- 30%

Inference

Please put test images into yolov3/data/samples result will appear in output/ Run detect.py to apply trained weights to an image

Performance

Run test.py to validate

1. Red-Round Traffic Sign Detection

(from TSRD dataset of NSFC)

127 epoches

2. Text Detection 50 epoch training details :

num of epoch	resolution
1-15	416x416
16-35	608x608
36-50	608-960(multi-scale)

Issues and future work

1. Multi-GPU training
2. Text detection and classification
3. Combine with specific type of sign