Last July, Xilinx announced the first-ever Xilinx Adaptive Computing Challenge. Run in partnership with Hackster.io, the challenge tasked developers with finding creative workload acceleration solutions using the Vitis Unified Software Platform and Vitis AI. Now, six months later, Xilinx has announced the winners of that inaugural competition.
The Adaptive Computing Challenge consists of three categories: Intelligent Video Analytics (using the Zynq UltraScale+ MPSoC ZCU104 Evaluation Kit), Adaptable Compute Acceleration (using the Alveo U50 Accelerator Card) and Adaptive Intelligence of Things (using the Avnet Ultra96-V2 Development Board). The challenge is awarding prizes to nine projects, three for each of the categories. In a press release, Xilinx announced the winners, who received first-, second- and third-place prizes of $10,000, $5,000 and $3,000, respectively.
Adaptable Compute Acceleration
- Acceleration of Binary Neural Networks using Xilinx FPGA by Raul Valencia: Exploit Xilinx FPGA’s hardware to train neuroevolved binary neural networks, then solve Reinforcement Learning problems.
- Covid4HPC – A fast and accurate solution for Covid detection by Dimitrios Danopoulos: Detecting Covid-19 from X-Ray images using CNNs on cloud FPGAs.
- ThunderGP: HLS-based Graph Processing Framework on FPGAs by Xinyu Chen: ThunderGP enables data scientists to enjoy the performance of FPGA-based graph processing without compromising programmability.
Intelligent Video Analytics
- Facemask Detector by Victor Altamirano: FPGA-based system that monitors facemask use through artificial intelligence, includes a thermometer and facemask dispenser.
- Automatic fall detection for elderly people by Jinin K Jose, Nevil Shah, and Rohin Kumar: Human falls is a major reason for deaths in elderly people. It can be prevented by an automatic fall detection, and alert system.
- Checkout So Easy – Real-time Smart Retail System For FPGA by Team MAAX: Deploy an object detection model on DPU to build a system that can show detected commodities in VCU decoded video or images from camera.
Adaptive Intelligence of Things
- Quad96 by Ussama Zahid: Quadcopter control and pole balancing using Deep Reinforcement Learning and Hand Gestures on Ultra96
- Hardware Accelerated Real-time Perception in 3D (HARP-3D) by Sambit Mohapatra: End-to-end demonstration of 3D object detection in LiDAR point clouds using a deep neural network running on the ULTRA96V2.
- LAMP-FPGA: Accelerating Time Series Similarity Prediction by Amin Kalantar and Philip Brisk: Predicting similar patterns in time series data on Ultra96-V2 FPGA board
“We received over 70 unique global project submissions,” Xilinx wrote. “Every participant showcased their expertise through their innovative design solutions. From the acceleration of binary neural networks to solve reinforcement learning challenges to an FPGA-based system that monitors facemask use through AI and automatic fall detection and alert system for older adults, the creativity and thought process dedicated to the challenge was truly on display in each innovative design solution.”
To learn more about the winners, follow this link.