The field of RADAR technology holds a lot of potential in building good applications like the Traffic Management system, while also preserving the privacy and mitigating the external environmental factors. Working with this technology, has allowed us to expand our application and try to design and build innovative products.
What it does
Our system uses state-of-the-art Computer Vision model to analyse and process the data gathered from the RADAR sensors and allow us to locate and classify various objects of interest such as cars and people. Our system can be used for various applications such as Traffic Management, Autonomous driving, and other similar applications which require some form of spatial information.
How we built it
We used the Pytorch framework and the Python language to first pre-process our data, and then design and train our model on this processed dataset. We researched some of the existing systems and architecture in order to understand and define our model as per our requirements and objective.
Challenges we ran into
Firstly, the data provided to us was limited and quite simple. We had to understand and then come up with techniques like Data Augmentation and data transformation to help improve our data. Secondly, even though we chose an established model, we weren't able to find proper pre-trained weights and thus fine-tuning was not possible in the given time. Lastly, we had to adopt the state-of-the-art model in order to make it compatible with our problem statement and dataset. Unfortunately we weren't able to reach convergence as we need more time to do Hyper Parameter search.
Accomplishments that we're proud of
In the short time of our hackathon, we got a good understanding of how RADAR technology works, seeing its advantages, disadvantages and its application. Working with such a tricky data allowed to explore any image processing and solutions that we could consider. We also managed to build, debug and start training a model in the short time. This itself was a challenge and we were quite happy to get some results at the end.
What we learned
We learnt about the RADAR technology and about how we can integrate the technology with some of the state-of-the-art techniques in Computer Vision. We also tried to understand what types of innovative applications could be derived from our work and how we could essentially use our product for the betterment of the society. These last two days also gave us a chance to interact with people from different backgrounds and profiles. We learned a lot from working together as team while being physically separated.
What's next for fahrRAD
We would love to work further on our model and try to tackle some of the challenges that we couldn't solve in this time frame. Apart from that, we also would love to try to look into of the interesting applications, the Traffic Management systems, and see how we can successfully integrate this technology of ours into this application.