Inspiration
Basically, an extension of Tracking.js features demo.
What it does
Select an image to use as query. Best matched image is selected from a database of images. All images were lifted from NASA gallery. NOTE: sometimes results gets messed up, just try again by selecting another image (or refresh the page).
How I built it
Tracking.js does most of the work. Features are extracted from query image. For each image in database, features are extracted and compared with query image's features. A score is computed and image with the best score is the winner.
Challenges I ran into
Biggest challenge was finding a good matching algorithm (pre-processing and scoring). I used a simple average of confidence scores.
Accomplishments that I'm proud of
Developing image search in a single page with plain HTML and Javascript.
What I learned
Tracking.js is simpler and more lightweight than OpenCV and similar computer vision libraries.
What's next for (EISE) Easy Image Search Engine
Looking forward to developing more web apps that use computer vision.
Log in or sign up for Devpost to join the conversation.