sghoshcvc / visualAttentionText

Implemementation of the paper Visual Attention Model for text recogntion https://arxiv.org/abs/1706.01487

TextProposals

1. Compilation

Requires: pyCaffe and Theano

The path for caffe needs to be provided in the file <model\.paths>

2. Download the DictNet pre-trained CNN model, protofile for dictnet and LSTM weights and parameters

3. Execute

Executing the following command:

./recognize_words_v1

process word images from /image subdirectory, result will be placed in <outputfile> one per image:

image_file_name, <transcript>