Inspiration
Radar is a well-known technology which has been recently gaining attention again because of advances in neural network-based classification. Radar can be applied to traffic detection because of its affordability and anonymity. In DarthRadar, we explore various machine learning methods to detect, count and classify targets: cars, pedestrians and trucks.
What it does
Radar-based detection, counting and classification of pedestrians, cars and trucks
How we built it
0) Data analysis 1) Peak detection, bounding box calculation, decision tree classifier 2) CNN radar target classifier 3) R-CNN end-to-end detector and classifier
Challenges we ran into
- a small and noisy dataset with unequally represented classes
- a faulty peak detection algorithm
Accomplishments that we're proud of
- working as a team until the end
- implemented three models that work on the given task in only three days
What we learned
- a lot about Computer Vision
- pipeline of Object detection and classification
- working under time pressure
What's next for Infineon Radar Challenge
- getting meaningful results from R-CNN
- improving peak detection
- utilizing a larger dataset
Built With
- keras
- python
- pytorch
- scikit-learn
- tensorflow
Log in or sign up for Devpost to join the conversation.