GianlucaMancusi / Paintings-Recognition-VCS

Computer vision and cognitive approachesin the Galleria Estense Museum

Authors: Gianluca Mancusi, Daniele Manicardi, Vittorio Pippi

The project aims to provide an application capable of processing paintings in images and videos, taken from the Galleria Estense in Modena.

Installation

1. (optional for the retrieval) Save a copy of the paintings_db in the following directory: dataset\paintings_db

2. Download the weights of YOLO from here: https://pjreddie.com/media/files/yolov3.weights and save the file in the yolo directory.

3. You have to install PyTorch and torchvision.

   A Windows only example of how to install PyTorch:

   pip install torch===1.5.1 torchvision===0.6.1 -f https://download.pytorch.org/whl/torch_stable.html

4. Install the requirements.txt in your virtual environment

   pip install -r requirements.txt

5. (optional) Download the weights of the U-Net and save them wherever you want. You need to log in with the institutional account (UNIMORE). From this URL: https://drive.google.com/drive/u/2/folders/1J1imEqytdpz8P9lT2gBuB75a2rnP6HDo

How to test the project

To start the project just run the python gui.py file, which will launch a web GUI from which you can test a pre-packaged image file or you can upload a new image or video file and test it.

Please note: the video output will be in uploads\videos\outputs It is better not to compute too big (40MB) files or very long video.

U-Net test:

Run predict.py in the U-Net directory and give the weights you would like to test, the input and the output filename. predict.py --model MODEL.pth --input IMAGE_FILENAME