💡 Inspiration
Around 15,000 accidents occur every day, the bulk of them are caused by the driver's recklessness. So we want to build Safe Drive 🚗 to avoid accidents and give the driver better and safe driving.
❓ What it does
Our project aims to reduce the number of accidents that occur on a daily basis by developing a collision avoidance system that will take into account a variety of inputs such as a live camera feed from a camera mounted above the car's windshield, distance between objects using a distance sensor, and CNN (YoloV5 architecture) for object detection and classification for classifying the types of objects the car must avoid.
🔨 How we built it
We use Webot Simulator to simulate our modal in real time.
we use CNN to build the semi-automatic modal for collision avoidance, where we use deep learning.
we use VOLOV5 architecture for object detection for better performance.
we also used COCO datasets for helping in object detection.
🏃♂️ Challenges we ran into
Learning Webot simulator as we have never worked with this simulator
Figuring out the steering angle and the speed variations
Some objects(cars, truck, person) were not detected, So we make the modal more precision
Finding the appropriate dataset
🏆 Accomplishments that we're proud of
We are proud on ourselves that we learned Webot Simulator and make this modal within time limit.
This is our first CNN project, We have never used neural network before, but this comes out pretty good, and we are proud on ourselves that we build this project.
📚 What we learned
We learned how to use Webot Simulator
We learned neural networks and how to build modal using it.
We learned how to make our own modals using neural network.
We also learn object detection and its best approaches like YOLOV5 architecture and implement in our project.
⏭ What's next for Safe Drive
We would like to implement this project in real world with a real car.
In the future, we also want to improve our modal for better performance.
Log in or sign up for Devpost to join the conversation.