For a deeper understanding the text detection alg called PixelLink, so I implemented the network model in keras just for testing.
- python 3.6
- keras 2.2.0
- opencv 3.4.2
The original PixelLink model trained with tensorflow, I converted the TF CHECKPOINTs(only PixelLink + VGG16 4s) into keras H5 weights, and you can download from here. Due to limited time, I've just implemented the model structure and testing process.
any suggestions are welcomed!