11 papers
SAGA: Learning Signal-Aligned Distributions for Improved Text-to-Image Generation
Paul Grimal, Michaël Soumm, Hervé Le Borgne +2
State-of-the-art text-to-image models produce visually impressive results but often struggle with precise alignment to text prompts, leading to missing critical elements or uninten…
The Deleuzian Representation Hypothesis
Clément Cornet, Romaric Besançon, Hervé Le Borgne
We propose an alternative to sparse autoencoders (SAEs) as a simple and effective unsupervised method for extracting interpretable concepts from neural networks. The core idea is t…
Reliable and Reproducible Demographic Inference for Fairness in Face Analysis
Alexandre Fournier-Montgieux, Hervé Le Borgne, Adrian Popescu +1
Fairness evaluation in face analysis systems (FAS) typically depends on automatic demographic attribute inference (DAI), which itself relies on predefined demographic segmentation.…
CaMiT: A Time-Aware Car Model Dataset for Classification and Generation
Frédéric LIN, Biruk Abere Ambaw, Adrian Popescu +3
AI systems must adapt to evolving visual environments, especially in domains where object appearances change over time. We introduce Car Models in Time (CaMiT), a fine-grained data…
MVAT: Multi-View Aware Teacher for Weakly Supervised 3D Object Detection
Saad Lahlali, Alexandre Fournier Montgieux, Nicolas Granger +2
Annotating 3D data remains a costly bottleneck for 3D object detection, motivating the development of weakly supervised annotation methods that rely on more accessible 2D box annot…
Explaining How Visual, Textual and Multimodal Encoders Share Concepts
Clément Cornet, Romaric Besançon, Hervé Le Borgne
Sparse autoencoders (SAEs) have emerged as a powerful technique for extracting human-interpretable features from neural networks activations. Previous works compared different mode…