Autonomous Robotics Research Center's Infrared Tracking Challenge

Autonomous Robotics Research Center's Infrared Tracking Challenge

Autonomous vehicles (AVs) today are defying the odds as they continue to disrupt the world as we know it with ever-greater advancements and capabilities. Autonomous ground vehicles can already understand their environment by applying deep-learning-based algorithms on top of perceptual information acquired from their sensors. The next generation of AVs will likely be expected to navigate low-light environments, proving something of a challenge to the photometric visual perception-based tracking algorithms frequently used in such scenarios.

For the uninitiated, photometric or RGB sensors are visible camera sensors comprising an imager that collects visible light (400~700 nm) and converts it into an electrical signal, then organizes that information to render images and video streams. Arguably, this could be an issue in ultra-low light environments. One solution to overcome this obstacle is to use thermal imaging sensors that quantify heat data instead of light.
To learn more and explore best-in-class approaches, Technology Innovation Institute’s Autonomous Robotics Research Center (ARRC) launched an Infrared Tracking Challenge to track objects in low-light conditions on December 19, 2022.

This crowd-sourcing challenge – in partnership with Hero X, a platform to facilitate such challenges to address opportunities or problems, is inviting innovators, technical experts, research institutes, and university students from around the world to develop approaches to efficiently track objects in the dark, in both structured and unstructured environments – the objects in question could include pedestrians, buggies, and motorcyclists, to name a few.

The participants can leverage datasets with content like TII’s sample dataset and work with color-to-thermal domain adaptation techniques - considering the scarce public availability of annotated thermal video sequences. Submissions are not limited to learning approaches – rather, classical computer vision approaches or a combination of both classical and AI-based approaches are welcome. TII will then evaluate the submissions based on common tracking metrics in the RGB domain. The nearly five-month-long challenge will conclude on May 17, 2023, with the announcement of three winners, who will win cash rewards of US$40,000 (first place), US$20,000 (second place), and US$15,000 (third place).

The final standalone application should be capable of running on Embedded AI Computer in real-time. For submission format details and specific information on the timeline of the Infrared Tracking Challenge, please visit May the best team win!