Logo Prediction
Published:
Created a solution to capture the logo of a company from images of products.
A system was used that captured all the keypoints from an image and matched them to keypoints from another image. A tool was built to create a collage of logos dynamically and compare this collage against any product to identify which logo it most strongly correlated with.
The matching logic:
- Compared logos relative to each other.
- Considered it a match if one logo matched with significantly more confidence than the others.
- If no strong match was found, or if it matched everything equally, it was flagged as a new brand logo.
- The logo was then cropped from the region with the most keypoints.