Welcome to ACID web page Presentation
Created by public researchers form the LaBRI (France)
Work supported by Aquitaine Région (DocHuman project) and the Bordeaux University (France)
ACID is an interactive software for visualizing and classifying big document image databases
Most of document image analysis systems (for a classification or a retrieval task) provide a very basic visualization interface where images are displayed and browsed sequentially. When databases are composed of many thousands of document images that kind of basic interfaces are really difficult to handle for a user.
ACID allows to visualize and browse a document image database as a graph and provides many visual indicators to focus the operator attention on potential errors.
The way ACID works is fully detailed in a paper presented in ICDAR2015
#Install docker package
sudo apt-get install docker.io
sudo addgroup $USER docker #need to restart the session to be part of docker group
#Download ACID sources (SIZE MO)
wget http:///acid.labri.fr/data/acid.tar.gz
tar -xzf acid.tar.gz
#If needed, in the Dockerfile, replace the package "nvidia-331-updates" by the one installed on the host
#Compile ACID
#Between 10 minutes and 45 minutes according download speed and computer performances
cd ACID
./docker.sh build
#To Test ACID
./docker.sh run
Feel free to download our groundtruthed database (8.8 MO) used to test ACID. This database is provided by Gestform company. All images have been anonymized. You can freely use this database.
If you use this database, please cite this paper in the original paper
