This work is the second part of the project - "Advancing Analysis and Cognitive Technologies for Art & Cultural Heritages", the first part work is related to an analytic framework, called DeepArt, details can be found in this Page.
In this work, we want to do an comprehensive research on high performance computing system to support our analytic framework. The research can cover three directions:
1. Investigate deep hashing techniques for fast large scale image retrieval.
2. Investigate deep network architectures for computing resources limited devices (mobile, wearable device), in which an interactive system associated with external equipment (Beacon) can be implemented.
3. Estimate and tune hyperparameters of models using more efficient methods (Bayesion method).

System Architecture

We implemented a real-world system for digital artworks retrieval and annotation. (a) is the coordinate server which contains webserver, image server and database, (b) is the online computational unit which can encode query image real-time, (c) is the offline computational unit which can encode images, cluster feature vectors and rank result.

More details are coming soon...

Live Demo

Based on the proposed framework and the collected large scale dataset (Art500K), we implement an artwork retrieval and annotation system which can be tried out by clicking the demo button:

Access Live Demo

Some Results



[Bibtex] [PDF] [Code] [Presentation] [Poster]


This work is supported by the HKUST-NIE Social Media Lab.