dhsegment.md

title	dhSegment
author
lang	en
documentclass	article
tags	software ocr htr
abstract

dhSegment

https://dhsegment.readthedocs.io/en/latest/

Prerequisites

Training

• nvidia GPU server with 6GB RAM au moins, avec CUDA et cuDNN.

Using

Data

Images / labels pairs, with the same name, and different extension (e.g., .png and .txt).

Demo

Installation

Prequisites: python3.6, git,

# Creating virtualenv
virtualenv env -p /usr/bin/python3.6
source env/bin/activate
# Installing
pip3 install git+https://github.com/dhlab-epfl/dhSegment

# Missing deps?
pip3 install scikit-image==0.13.1

Demo

(Demo)[https://dhsegment.readthedocs.io/en/latest/start/demo.html]

git clone https://github.com/dhlab-epfl/dhSegment.git
cd dhSegment/demo/
wget https://github.com/dhlab-epfl/dhSegment/releases/download/v0.2/pages.zip
unzip pages.zip
cd ..
cd pretrained_models/
python download_resnet_pretrained_model.py
cd ..
# Train a model, or don't, and download an existing model (as follows)
cd demo/
wget https://github.com/dhlab-epfl/dhSegment/releases/download/v0.2/model.zip
unzip model.zip
cd ..
# Modification
python demo.py

• modifier boxes_detection.py, l. 28
• l. 28, retirer _, avant contours.