We are building a system for a chicken delivery hub in Ukraine. They face the problem of stealing goods while weighing boxes and loading them to trucks.
The solution of the problem is automated confirmation of boxes count, located on the pallette, which is being weighted. In this case, employees are not able to put into the system wrong number of boxes, which they did to deceive the system and steal goods.
This system consists of:
- Cameras, established in a viewpoint of weigher, which has a view on the palette from all angles.
- Deep learning model based on pre-trained ResNet34 and handled by PyTorch, which is trained to detect boxes on an image and can count number of boxes on each.
- Analytical software, which use model, mentioned about to detect boxes and count an exact number of boxes on each palette.
- Software that connects analytical system with company's PIM