An application for colorizing and recovering old black&white photos and videos

python3 main.py --image_file test-image.jpg --output_file recovered-image.jpg

Full options list is as follows:

usage: main.py [-h] -i IMAGE_FILE (-o OUTPUT_FILE | -s) [-d DEVICE] [-v] [-S]

optional arguments:
  -h, --help            show this help message and exit
  -i IMAGE_FILE, --image_file IMAGE_FILE
                        an input image file
  -o OUTPUT_FILE, --output_file OUTPUT_FILE
                        an output image file
  -s, --show            display an output image on the screen instead of
                        saving into file
  -d DEVICE, --device DEVICE
                        target device for infer: CPU, GPU, FPGA, HDDL or
                        MYRIAD; defaults to CPU
  -v, --verbose         enable display of processing logs
  -S, --smoothing	use smoothing filter

For the first run one needs to:

1. Install Intel OpenVINO with Python3
2. Run (OPEN_VINO INSTALL PATH)/bin/setupvars.sh
3. Run init.sh shell script (for Linux/MacOS), which does:
   1. Setup a python virtualenv
   2. Install requirements.txt
   3. Create ./data/models folder
   4. Download necessary models

source ./init.sh

Formerly recommend to work within local virtualenv

• main.py -- main project file
• helper/
  • colorization.py -- colorization network invocation
  • superres.py -- super-resolution network invocation
  • cv_tools.py -- helper functions on OpenCV
  • common.py -- global variable definitions
• data/ -- generated after the init script is being executed
  • models/ -- expected to contain models automatically downloaded by Intel OpenVINO ModelDownloader and processed by ModelOptimizer down to the IR format
• requirements.txt -- pip virtualenv requirements
• LICENSE -- license information

• Create an init shell script for Windows (?)
• Command line args parsing + help output
• Colorization network execution
• Super-resolution network execution
• cv_tools implementation