rathorology / fots_tf1.150

Vehicle license plate detection on cropped images of license plate

Fast Oriented Text Spotting with a Unified Networks

Introduction

This is an use of FOTS: Fast Oriented Text Spotting with a Unified Network

Install

• Python3.5
• tensorflow 1.12.0
• OpenCV

pip install -r requirements.txt
git clone -b dev https://github.com/Pay20Y/FOTS_TF.git

Extract images to one location

python3 uniform_images.py

Make annotation for training in FOTS format

python3 make_annotation.py

Used pretrained model to train

SynthText 6-epochs pretrained model can be found here

Train

python3 main_train.py --gpu_list='-1' --learning_rate=0.0001 --train_stage=2 --training_data_dir=data --training_gt_data_dir=annotation

Test

python3 main_test.py --gpu_list='-1' --test_data_path=/path/to/testset/ --checkpoint_path=checkpoints/

Align results with ground truth for train and test data

python3 final_output.py

Evaluate model on test

python3 evaluation.py

Test Accuray

72.21%

Metric for evaluation

Percentage match of strings

fots_tf1.150's Issues

Data Directory

Can you please mention about "How to set data directory?

Thank you.