Computer Vision - Museum Painting Detection

Authors: Marco Cagrandi, Alessio Ruggi and Daniele Lombardo.

GitHub Repository: https://github.com/Spucis/CV_PaintingDetection

The file cv_project.py is the entry point of the application. It contains the video selection and it is where you can change the step of the program. The file conf.json is the configuration file and it's were you can change the input directory and (optionally) all the paths of the directories of the project.

• input directory contains all the provided inputs such as the images, the videos, the .csv file and the map
• output directory contains all the produced outputs (e.g. the videos)
• source contains the source files:
  • detection_utils.py contains all the utilities to perform the detection phase (e.g. edge detection and connected components detection)
  • painting_manager.py contains all the functions needed to fulfill the requested work
  • globals.py contains all the global definitions and the utilities used to manage the input video itself (such as opening and closing the video).

An output file output_details.json is provided: it summarize all the information provided by the program.

Each step frames the program will stop and process the frame, producing a modified frame mod_frame and writing it on the disk. In cv_project.py is possible to enable the output of the details with the option json_output_details, enable the segmentation of both statues and detected paintings with the option en_segmentation, and modify the step.

Instead, inside the painting_manager.py file and paint_detection(...) function you can manipulate other options:

• verbose option which will enable more output over the standard output
• show_details will stop the execution to let you see the detected ROI, the rectified attempt and the proposed selected painting in the DB
• otsu_opt_enabled will enable the otsu_optimization, as discussed in the report

https://drive.google.com/open?id=1Na6uMDc_ST1179GNjGjJ5ywywlBoYXNj