arXiv subjects

Urszula Czerwinska

Publications and source records attributed to Urszula Czerwinska.

3 recordsLinked to original sources

arXiv2026

MONET: A Massive, Open, Non-redundant and Enriched Text-to-image dataset

Training large text-to-image models requires high-quality, curated datasets with diverse content and detailed captions. Yet the cost and complexity of collecting, filtering, deduplicating, and re-captioning such corpora at scale hinders open and reproducible research in the field. We introduce MONET, an open Apache 2.0 dataset of approx. 104.9M image--text pairs collected from 2.9B raw pairs across heterogeneous open sources through successive stages of safety filtering, domain-based filtering, exact and near-duplicate removal, and re-captioning with multiple vision-language models covering short to long-form descriptions, and further augmented with synthetically generated samples. Each image is shipped with pre-computed embeddings and annotations to accelerate downstream use. To validate the effectiveness of MONET, we train a 4B-parameter latent diffusion model exclusively on it and reach competitive GenEval and DPG scores, demonstrating that our dataset lowers the barrier to large-scale, reproducible text-to-image research.

cs.CV↗

arXiv2025

Benchmarking Image Embeddings for E-Commerce: Evaluating Off-the Shelf Foundation Models, Fine-Tuning Strategies and Practical Trade-offs

We benchmark foundation models image embeddings for classification and retrieval in e-Commerce, evaluating their suitability for real-world applications. Our study spans embeddings from pre-trained convolutional and transformer models trained via supervised, self-supervised, and text-image contrastive learning. We assess full fine-tuning and transfer learning (top-tuning) on six diverse e-Commerce datasets: fashion, consumer goods, cars, food, and retail. Results show full fine-tuning consistently performs well, while text-image and self-supervised embeddings can match its performance with less training. While supervised embeddings remain stable across architectures, SSL and contrastive embeddings vary significantly, often benefiting from top-tuning. Top-tuning emerges as an efficient alternative to full fine-tuning, reducing computational costs. We also explore cross-tuning, noting its impact depends on dataset characteristics. Our findings offer practical guidelines for embedding selection and fine-tuning strategies, balancing efficiency and performance.

cs.CV↗

arXiv2015

DeDaL: Cytoscape 3.0 app for producing and morphing data-driven and structure-driven network layouts

Visualization and analysis of molecular profiling data together with biological networks are able to provide new mechanistical insights into biological functions. Currently, high-throughput data are usually visualized on top of predefined network layouts which are not always adapted to a given data analysis task. We developed a Cytoscape app which allows to construct biological network layouts based on the data from molecular profiles imported as values of nodes attributes. DeDaL is a Cytoscape 3.0 app which uses linear and non-linear algorithms of dimension reduction to produce data-driven network layouts based on multidimensional data (typically gene expression). DeDaL implements several data pre-processing and layout post-processing steps such as continuous morphing between two arbitrary network layouts and aligning one network layout with respect to another one by rotating and mirroring. Combining these possibilities facilitates creating insightful network layouts representing both structural network features and the correlation patterns in multivariate data. DeDaL is the first method allowing to construct biological network layouts from high-throughput data. DeDaL is freely available for downloading together with step-by-step tutorial at http://bioinfo-out.curie.fr/projects/dedal/.

q-bio.QM↗