We have a database of "Reference images" representing product boxes. For example, an image could be a 1000x1500 pixel JPG cropped around the product's face. We also have a large folder of pictures of product boxes on shelves. All images have a reasonable resolution.
We require the software to run thru the folder, load each shelve image, and match the reference images from our database, then return the bounding boxes of each match. We need this software to be very reliable, we want it to use multiple validation methods (found features, OCR the box's text, compare colors, detect the boxes WxH ratio...). The software will output its detections in an SQL database.
We expect this software to be programmed using OpenCV, and use multi-thread for speed although we understand there will be latency, the software does not need to process images in real-time. We are looking for an experienced programmer who has successfully performed projects like this before, and we want the candidate to be able to demo previous work, and give us a precise description of what tools and modules, and algos he would use for this project.