Congratulations to Oleksii Shashko, a PhD student of the AI Department, and his academic supervisor Anastasiia Deineko on their publication in a first-quartile journal with international co-authors.
The article “Predicting High-precision Depth on Low-Precision Devices Using 2D Hilbert Curves” focuses on applying deep neural networks for dense depth prediction (DNN) on low-end devices. For on-device deployment, the researchers used standard quantization methods with post-processing that reconstructs depth from the components of a Hilbert curve predicted with low precision.
Large-scale experiments demonstrate that the proposed method increases depth prediction accuracy by up to three bits with minimal computational cost.
We wish Oleksii continued success in his scientific work!