Project Description
The retail-centric app is an intelligent Python-developed solution for stock management in a supermarket. The AI agent, called SHAI, ensures a constant flow of products on shelves by analyzing and recognizing products and empty spaces.
This is done through a planogram which is realized with video cameras aligned on the opposite shelves. It is another great example of automation that reduces human error and repetitive tasks.
Technical Description
- Uses two Models, one pre-trained model to recognize cans by shape and one was trained in the contest night to recognize the actual product. For training, we took 1000 photos. The first model helps to recognize a beer can from empty spaces and also crops individual cans images recognized for improving product recognition by the second model. On testing data, a 99% accuracy was reached for 5 different types of beer cans.
- Models were created using Keras and TensorFlow.
Mobile app allows taking pictures, sending it to Backend to interpret and displays a report that shows: Total products displayed, number of each type of product recognized, empty spaces and mismatched places in the shelf. - Backend handles the communication between Mobile app and AI Model by receiving the photo from the Mobile app and send it to the Model for interpretation and taking the interpretation result from Model and send it to the mobile app as a JSON object.
Project beneficiaries
- Retail industries
- Stores
- Horeca industry
