269 citations · 291 across the 10 of their papers we have counts for
13 papers
IDeaL: Data-Free Multi-Teacher Distillation via Improved Dead Leaves
Feyza Yavuz, Mert Bülent Sarıyıldız, Diane Larlus
Multi-teacher distillation has emerged as a way to combine complementary teacher models into a single student model that exhibits the strengths of all its teachers. The student is…
Compressing History into Memory: Distilling Transformers into Recurrent Transformers
Philippe Weinzaepfel, Christian Wolf, Bülent Mert Sariyildiz +3
Transformers are AI's workhorse but their computational cost becomes prohibitive when processing long sequences. We target long-horizon streaming vision and robotics applications,…
Task Alignment: A Simple Proxy for Practical Model Merging Across Diverse Vision Tasks
Pau de Jorge, César Roberto de Souza, Björn Michele +5
Efficiently merging several models fine-tuned for different tasks, but stemming from the same pretrained base model, is of great practical interest. Despite extensive prior work, m…
Kinaema: a recurrent sequence model for memory and pose in motion
Mert Bulent Sariyildiz, Philippe Weinzaepfel, Guillaume Bono +2
One key aspect of spatially aware robots is the ability to "find their bearings", ie. to correctly situate themselves in previously seen spaces. In this work, we focus on this part…
DUNE: Distilling a Universal Encoder from Heterogeneous 2D and 3D Teachers
Mert Bulent Sariyildiz, Philippe Weinzaepfel, Thomas Lucas +3
Recent multi-teacher distillation methods have unified the encoders of multiple foundation models into a single encoder, achieving competitive performance on core vision tasks like…
UNIC: Universal Classification Models via Multi-teacher Distillation
Mert Bulent Sariyildiz, Philippe Weinzaepfel, Thomas Lucas +2
Pretrained models have become a commodity and offer strong results on a broad range of tasks. In this work, we focus on classification and seek to learn a unique encoder able to ta…
Weatherproofing Retrieval for Localization with Generative AI and Geometric Consistency
Yannis Kalantidis, Mert Bülent Sarıyıldız, Rafael S. Rezende +3
State-of-the-art visual localization approaches generally rely on a first image retrieval step whose role is crucial. Yet, retrieval often struggles when facing varying conditions,…
Fake it till you make it: Learning transferable representations from synthetic ImageNet clones
Mert Bulent Sariyildiz, Karteek Alahari, Diane Larlus +1
Recent image generation models such as Stable Diffusion have exhibited an impressive ability to generate fairly realistic images starting from a simple text prompt. Could such mode…
No Reason for No Supervision: Improved Generalization in Supervised Models
Mert Bulent Sariyildiz, Yannis Kalantidis, Karteek Alahari +1
We consider the problem of training a deep neural network on a given classification task, e.g., ImageNet-1K (IN1K), so that it excels at both the training task as well as at other…
Concept Generalization in Visual Representation Learning
Mert Bulent Sariyildiz, Yannis Kalantidis, Diane Larlus +1
Measuring concept generalization, i.e., the extent to which models trained on a set of (seen) visual concepts can be leveraged to recognize a new set of (unseen) concepts, is a pop…
Hard Negative Mixing for Contrastive Learning
Yannis Kalantidis, Mert Bulent Sariyildiz, Noe Pion +2
Contrastive learning has become a key component of self-supervised learning approaches for computer vision. By learning to embed two augmented versions of the same image close to e…
Learning Visual Representations with Caption Annotations
Mert Bulent Sariyildiz, Julien Perez, Diane Larlus
Pretraining general-purpose visual features has become a crucial part of tackling many computer vision tasks. While one can learn such features on the extensively-annotated ImageNe…
Key Protected Classification for Collaborative Learning
Mert Bülent Sarıyıldız, Ramazan Gökberk Cinbiş, Erman Ayday
Large-scale datasets play a fundamental role in training deep learning models. However, dataset collection is difficult in domains that involve sensitive information. Collaborative…