Inspiration
Few days back when self-driving car killed a pedestrian, I was wondering that we need some kind of alert system for the car to let it know when to stop automatically. My idea was to detect vehicles and before they hit each other, give a warning to each car saying - 'you are going to collide' .
What it does
Currently due to limitations of system and CPU, I was able to detect objects using the live-feed from camera.
How I built it
I used tensorflow, scikit, opencv to desing this application
Challenges I ran into
GPU: Tensorflow CPU is too slow. and it keeps on crashing system if you try to detect things from a video. CUDA: Only supports NVIDIA graphics.
Accomplishments that I'm proud of
It detects live fee camera objects with good accuracy and detects facial expressions.
What I learned
A lot of system dependencies
What's next for what is this!
Try to make it in the system having GPU. The algorithm can be improved a lot.
Built With
- machine-learning
- python
- scikit-learn
- tensorflow
Log in or sign up for Devpost to join the conversation.