GCCNet

Code for the corresponding neurocomputing paper.

Description

This repo contains the main application of the GCCblock on text detection and also the application on CoQA(a nlp task)

We will see if we can release the ensembler, but at least here is the code for single models.

Paths

This repo has 3 improtant paths, which are $DATASET,$CODEROOT, and $MODELROOT.

$DATASET holds training and evaluation datasets.

$CODEROOT holds the code.

MODELROOT holds the trained model and evaluation results.

By default, these paths are set to:

$DATASET: /home/username/pubdata/datasets/testing_sets
$CODEROOT: /home/username/cat/project_tf_family
$MODELROOT: /home/username/cat/project_uniabc_data

You may want to change them in utils/libpath.py, note the name can be different.

Trained models

https://drive.google.com/drive/folders/1ZFVrKlKAYQO77cIIzcTYW0Iyeni2qkwe?usp=sharing

You need to put these two folders into MODELROOT

Evaluation

Run project_easts/tester.py

About

Code for the corresponding neurocomputing paper GCCNet: Grouped Channel Composition Network for Scene Text Detection.