Roadside Recognition Using HLS Hackathon Podcast

With only 2 weeks remaining, the hackathon leaderboard has shifted to picojoules to capture the thin margins separating the top teams, where the reigning leader sits at an ultra-efficient 9.2nJ per inference. Competitors must balance strict area, speed, and accuracy rules while fine-tuning parallelism and precision to break into the top 10 before winners are announced at Edge AI in Taipei on October 22nd.

The finish line is in sight, but we are not there yet, 4 weeks to go! This episode analyzes the current wave of submissions and celebrates the incredible progress of participants. We have seen energy consumption drop from thousands of NanoJoules all the way down to the double digits. Tune in as we explore the winning strategies emerging across the leaderboard. If you are struggling to find the right trade-off between performance, design area, and accuracy, listen in for our latest tips.

This episode dives back into the heart of the Roadside Recognition Hackathon and the quest for ultra-efficient edge inferencing. Revisiting the core requirements: accuracy, performance, area and design consumption. Looking at how teams are balancing these pillars as they master the HLS flow. The competition is officially heating up, especially since the current top entry is sitting at a remarkable 29 NanoJoules! The race is on and there are just six weeks left.

This episode explores how teams are balancing accuracy, performance, and area, giving you a look at what's working. Listen in to gain a competitive edge for your own hackathon journey!