An unofficial pytorch implementation of CRAFT: Character Region Awareness for Text Detection.
Part1: train your model on the synthtext dataset.
Part2: finetune your model on the ICDAR2015 dataset.(use some synthtext data here to make the model perform better)
After training on the synthtext dataset, input image to test.
After finetuning on the ICDAR2015 dataset, we can get the final result.
