Robustness Evaluation of Computer Vision in Traffic Videos

So, the aim of this project is to check how good computer vision tools are at spotting objects in traffic videos, especially when things get tricky in real life. The usual tests they do aren't really capturing all the crazy anomalies that can happen out there, which is a big deal as self-driving cars are getting popular. We're digging into how these smart algorithms handle things like bad weather, which can mess up their accuracy big time. Just because a system works great when the sun's out doesn't mean it won't fumble when it's rainy or foggy. Our research will also tinker with some fancy setups that use different methods, like checking these work using both cameras and sensors. We're looking at a bunch of ways this could be useful, like keeping an eye on traffic, following cars around, and tracking objects. The end goal? Make getting around safer and smoother for everyone.

• Role: Machine Learning Engineer
• Technologies: Tensorflow + Keras
• Semester: Summer 2023