We found the "Counting Cars" challenge interesting. Our team members brought to the table experience with:
- smart home
- smart city
- smart construction
- conversational A.I.
- applied A.I.
- cloud computing
What it does
The B.E.A.R. car counter relies on cloud computing, turning video (webcam feeds) into static video data, then searching for vehicle pictures within the video data. This challenge was on cars but we could easily also search for busses (known taxonomy category for MS Azure CV App image recognition) and on-demand train our model to also detect snowplows, trucks, etc. The B.E.A.R. car counter is also designed not to count stationary vehicles (e.g. on the side of the road, after a breakdown, or stuck in traffic) multiple times. Predictive modelling with an increase in vehicles leading to notifications to the road maintenance managers is another option. The same goes for detecting (and warning about) rapid changes in temperature, either based on computer vision and/or weather data feeds.
How we built it
During this weekend, we designed the solution and "build" the software architecture.
Challenges we ran into
Time constraints of team members due to academic, family, and medical reasons.
Accomplishments that we're proud of
- Understanding winter road maintenance from our research
- Taking MS Azure learning path modules on computer vision during this hackathon
- Our remote and asynchronous teamwork across two different EU countries (Austria, Germany)
What we learned
- In-depth insights into MS Azure Computer Vision App
- In-depth insights into Jina.ai Neural Search
- In-depth insights into Smart Roads
What's next for B.E.A.R. Car Counter
The MS Azure A.I. hackathon, most likely...