Inspiration
We live in a bustling commercial city where the distribution of goods and services is quite common. Unfortunately, so are accidents. Given our proximity to road accidents and observing that a significant portion of these incidents result from fatigue or distractions, we decided to create a solution aimed at detecting vulnerable situations for drivers. This solution provides recommendations based on what the computer vision system detects.
What it does
The project involves a web-based dashboard for monitoring delivery drivers within a specific area. Each driver is equipped with a camera integrated with a computer vision system that detects signs of fatigue, lack of concentration, and drowsiness. Based on these metrics, an autoencoding prediction model determines the likelihood of a traffic accident. Additionally, the web page allows users to visualize the routes taken by the drivers to select the optimal one.
How we built it
For the web interface, we used Next.js, React, CSS, and Geoapify for the mapping APIs. For intercommunication, we utilized Supabase. On the backend, the computer vision system was developed in Python, the neural network system was built with TensorFlow optimized with OpenVino, and the recommendation system was implemented with the help of Frida by Softtek.
Challenges we ran into
One of the main challenges was integrating a map API into the page and generating delivery routes. On the backend, creating a functional system to calculate the probability of a traffic accident without a specific dataset for driver fatigue was a significant challenge.
Accomplishments that we're proud of
We are proud of the dashboard system and its ability to provide comprehensive information about the drivers in a given area, the visual fatigue detection system, and the capability to predict the likelihood of a traffic accident.
What we learned
We gained valuable insights into implementing technologies provided by Softtekk, such as OpenVino and Frida, and their significant impact on optimization.
What's next for Zenti
We aim to enhance the accident prediction system with a more robust dataset, continually focusing on reducing traffic accidents caused by fatigue or distractions, which is the primary mission of Zenti.
Log in or sign up for Devpost to join the conversation.