/CTPN.caffe

Primary LanguageJupyter Notebook

Detecting Text in Natural Image with Connectionist Text Proposal Network

This repository is a caffe implementation of CTPN, including training code

Installation

  1. requirement: python2.7

Train

./train_ctpn.sh

Train loss

I1016 08:24:03.314466  7483 solver.cpp:228] Iteration 69980, loss = 0.0946643
I1016 08:24:03.314492  7483 solver.cpp:244]     Train net output #0: rpn_cls_loss = 0.0137189 (* 1 = 0.0137189 loss)
I1016 08:24:03.314512  7483 solver.cpp:244]     Train net output #1: rpn_loss_bbox = 0.0463248 (* 1 = 0.0463248 loss)
I1016 08:24:03.314517  7483 sgd_solver.cpp:106] Iteration 69980, lr = 1e-06

Test

cd ./tools
python do_test.py

TODO

  • training code
  • test code
  • evaluate on icdar2013 dataset
  • quantization INT8
  • low bit
  • rewrite convolution for bit computation