Benjamin Ramtoula

Benjamin is currently a postdoctoral researcher at the Mobile Robotics Group of the University of Oxford working with Daniele De Martini and Paul Newman. He pursued his DPhil in the same group as part of the CDT in Autonomous Intelligent Machines and Systems (AIMS).

Before moving to Oxford, he studied at EPFL, ČVUT, and the MISTLab of Polytechnique Montréal. He also spent some time at MSL in Stanford, and in Team CoSTAR at NASA JPL, working on the DARPA SubT Challenge.

His research focuses on using representations learned by foundation models to make robust systems.

Fantastic Features and Where to Find Them: A Probing Method to Combine Features from Multiple Foundation Models

ComBo combines features from different layers of multiple frozen vision foundation models using a lightweight, trainable adapter. It brings complementary representations together for downstream tasks and identifies which models contribute most, improving on individual backbones without fine-tuning the foundation models themselves.

Visual DNA: Representing and Comparing Images using Distributions of Neuron Activations

Visual DNA represents images and datasets through distributions of neuron activations in a frozen neural network. These compact representations allow comparisons that focus on selected visual properties, supporting dataset analysis, evaluation of synthetic imagery, and the selection of data for transfer learning.