Thomas Verelst

Senior Machine Learning Engineer @ Axelera AI

PhD (KU Leuven, Belgium) in computer vision and deep learning. Enabling deep learning and computer vision on edge devices. currently working on quantization, inference optimization and finetuning of LLMs and multi-modal Gen AI models.

First-Author Publications

BlockCopy: High-Resolution Video Processing with Block-Sparse Feature Propagation and Online Policies

Thomas Verelst, Tinne Tuytelaars ICCV 2021 [paper] [code]

SegBlocks: Block-Based Dynamic Resolution Networks for Real-Time Segmentation

Thomas Verelst, Tinne Tuytelaars TPAMI ECCV 2020 - Embedded Vision Workshop Best Short Paper Award received on August 28th 2020 | [TPAMI long paper] [ECCV short paper]

Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference

Thomas Verelst, Tinne Tuytelaars CVPR 2020 [paper] [Pytorch code] [YouTube 1-minute teaser]

Generating superpixels with deep representations

Thomas Verelst, Matthew Blaschko, Maxim Berman CVPR 2018 workshop on DeepVision: Beyond supervised learning (extended abstract) [paper]

Want to know more about conditional neural networks? Check out my reading list!

Other work

Processor Architecture Optimization for Spatially Dynamic Neural Networks

Steven Colleman, Thomas Verelst, Linyan Mei, Tinne Tuytelaars, Marian Verhelst 2021 IFIP/IEEE 29th International Conference on Very Large Scale Integration (VLSI-SoC) [paper]

“Can human behaviour, emotion and cognition be detected by computer vision techniques? A use case on student engagement”

Pieter Vanneste, José Oramas, Thomas Verelst, Tinne Tuytelaars, Annelies Raes, Fien Depaepe, Wim Van den Noortgate Mathematics 2021 journal paper [paper]