InShape
1️⃣ Original image (greyed)
3️⃣ If a region has a nice surface ratio, we extract (shape, color, text)
Cropped image
Size
Surface ratio
Shape recognition
Dominant color extraction
Optical Character Recognition (OCR)
Result
4️⃣ Other ROIs are just filtered out, they don't contain anything in terms of image processing
Cropped image
Size
Surface ratio
Shape recognition
Dominant color extraction
Optical Character Recognition (OCR)
Result