Skyhacks #1

Solution to ArcelorMittal - Computer Vision - UIC Wagon recognition - Computer-based solution challenge

Env requirements

• Node js (v8.12.0)
• npm (5+)

1. Install deps

npm i

2. Download footage

The footage should be located in /src/footage/Validation, /src/footage/Training, and /src/footage/Special. Special is for train 0_50 (or other).

If you put something to Special dir, please insert it there respecting the original directory structure: Special/0_{train}/0_{train}_{left|right}/{files}

3. Run csv generation

npm run verify-wagons

Grab your coffee, sit back, and relax. This takes approx. 20s-30s per train, so 60 trains gives 30 minutes.

Csv files lay in /src/verify-wagons/csv directory.

4. Run fisheye distortion removal

You need to modify the /src/fisheye/index.html file to adjust it to your needs. These are:

• const train = '63';
• const side = 'right';
• const containingDir = 'Special';
• const range = [0, 463];

Run npm run remove-fisheye, and wait until images download.

Pass the images to OCS processing

5. Then there is the OCR

You need to run ocr/process.py.

6. Pass UIC codes

• put txt files into /src/verify-wagons/data
• run npm run add-uics