EauDeData.github.io / research

Reading the archive,
verifiably.

An interactive collection — a doctoral thesis and the papers behind it, on historical document retrieval, verifiable vision, and decentralized analysis.

Adrià Molina · CVC-UAB 1 thesis · 5 papers click any cover to open
Doctoral Thesisspecial
thesis.html
open ↗
★ the whole thesis
Retrieving Document Images Under Domain-Specific Constraints

The full-length interactive thesis tying the strands together: controllability, distributed learning, temporal reasoning, and verifiable retrieval over historical documents.

scrollytelling5 chapters./thesis.html
Papers5
tdir.html
open ↗
Paper 01./papers/tdir.html
Composed Historical Image Retrieval by Modeling Temporal Representations
composed retrievaltemporal
date.html
open ↗
Paper 02./papers/date.html
Date Estimation in the Wild of Scanned Historical Photos: An Image Retrieval Approach
date estimationretrieval
checking.html
open ↗
Paper 03./papers/checking.html
Visual Model Checking: Graph-Based Inference of Visual Routines for Image Retrieval
verifiable retrievalmodel checking
dist.html
open ↗
Paper 04./papers/dist.html
Unapologetically Distributed: A Call for Decentralized Document Analysis
decentralizedfederated
fas.html
open ↗
Paper 05./papers/fas.html
Fetch-A-Set: A Large-Scale OCR-Free Benchmark for Historical Document Retrieval
OCR-freebenchmark