Artlabs Inc.

Product counting with AI

Overview ArtLabs team has developed a solution for item counting on the production line. The main problem of the client was the control of the production count. The company tried to use sensors to count the amount of bread on the conveyor, but the result was inaccurate, and had to use manual labor to count. Object Tracking Algorithm A computer vision solution has been proposed that counts bread with a single camera. As a result, it comes out quite cheaply, you just need a simple camera and computer. Having invested in training once, you can get a result of recognizing different types of bakery products. Also, machine learning can be tuned to reduce scrap. During our research, we tried to come up with a variety of algorithms including YOLO V3, RetinaNet, and others. In the end, we chose the EfficientDet algorithm due to the robust results and high-speed inference. UI & Dashboards In the end, we wrapped our software into a simple and user-friendly UI with access to dashboards and product line statistics. Our system was very simple to use and didn’t require human intervention for product counting. Moreover, it is very easy to fine-tune the existing model on new item sets in the inventory.