Project developed during the Transatlantic AI Hackathon 2022, Sustainable Sustainable Supply Chain DeepHack.
The idea has been previously studied by team TruckPooling in the Start-Up lab 2022 edition, held by CLab Trento, and presented in the Demo day on 26th May 2022.
The idea is to reduce green house gases emission of trucks implementing a semi-autonomous driving system, allowing the vehicles to drive close to each other and exploting a wind tunnel effect. This would lead to an heavy reduction of the air drag affecting the truck, reducing consumption and emisisons. Thus, this system would rise the efficiency of the vehicle, resulting in lower costs, and reducing pollution (for combustion-engine equipped trucks).
This project aims to build a simple plug-and-run system able to detect vehicle in front of the truck, exploting the OAK-D-Lite kindly provided by Luxonis.
Thanks to this device, we were able to rapidly combine a vehicle detection system based on artificial intelligence, and a stereo-camera system to detect distance and relative velocity with respect to the truck on which is mounted on.
- real-time run the algorithm in real time.
- recording run the algorithm on a registration done with OAK-D-Lite.
- truckpooling demo used during the Start-Up lab 2022 Demo day. It runs the same algorithm, but using a model to detect persons instead of vehicles.
First, you need to install the DepthAI library and its dependencies.
Install the requirements for the recording example:
python3 -m pip install -r requirements.txt
For detecting vehicle from a recorded video (place the video in the recordings folder)
python3 antTrail_recording.py
For detecting vehicle from from live camera stream
python3 antTrail.py