Package anomaly detection
We have developed algorithm utilizing OpenCV to match images and detect image anomalies, in given set of images.
Instead of choosing original reference image, algorithm develops own model on given data set.
Approach
We have tested multiple options and possible solutions to problem, our goal was to make solution as fast as possible. For that goal to be possible, we have discussed and tried to implement solution in more than one way.
Problems
QR code found in image has posed large problem, because even with efforts given to alignment we could not attach it properly and noise it generated was too large.
However, given other parts of our algorithm it can be possible to adapt it to solve problem with QR code on image.
Improvements
Most important improvement would be to align images a little better so noise gets reduced and to use cuda to compare and calculate diff values.
Log in or sign up for Devpost to join the conversation.