Surveys and Datasets

Continual Object Detection Benchmark

In Verwimp et al. Neural Networks 2023, we introduce CLAD, a novel benchmark for continual learning focusing on realistic distribution shifts. The benchmark, built on top of Huawei’s SODA10M autonomous driving dataset, comprises a classification (CLAD-C) and an object detection benchmark (CLAD-D). CLAD-C was built from a stream of chronological images, yielding smooth and realistic transitions in terms of domain (e.g. weather) and class distributions. CLAD-D is a domain incremental object detection benchmark, comprising 4 tasks associated with different environmental conditions. The CLAD benchmarks were released as part of the Self-supervised Learning for Next-Generation Industry-level Autonomous Driving Challenge and top proposed methodologies are discussed in our paper. Code and details are available here.

Continual Learning Survey

In De Lange et al. PAMI 2021, we carried out a detailed analysis and survey of continual learning methodology for classification tasks. We thoroughly evaluated 11 state of the art methods and the influence of different model components against 4 baselines, on 3 benchmark datasets. Code and details are available here.

Brain Parcellation Survey

In Arslan et al. NeuroImage 2017, we provide a large-scale systematic comparison of more than 30 state-of-the-art connectivity-driven, anatomical, and random parcellation methods. The preprint, as well as the code and data used for the survey are provided at the dedicated webpage.

Low-Grade Gliomas Database

We have released the database used during my PhD for tumour segmentation and analysis. This is a set of 210 FLAIR MRI of different patients suffering from a diffuse Low Grade Glioma. All the images provided were acquired in a clinical setting, with varying quality and evolution of the disease. It results in a very challenging database for image processing tasks.

The complete database can be found here (NIFTI format)