Garbage collection company faced a problem of plan execution. Garbage trucks should follow the plan of collecting a specific number of garbage containers in specific spots. But truck drivers don't always follow the plan and can miss containers or grab containers out of their schedule.
We developed a system that grabs data from cameras located on truck and checks where exactly container was loaded.
This system consists of:
- Cameras, established on a truck with a view on container loading spot.
- Deep learning model based on ResNet34, which is trained to detect garbage container loads using PyTorch.
- Analytical software, which uses script that detect loads, GPS data from monitoring system. Together these 2 dataflows can give exact information on when, where and how many containers were loaded.
- Software that connects analytical system with the company's PIM, where data from the system is synced with planning data. So management can track how garbage truck drivers are following their plans.