UNNA: A Unified Neural Network for Aesthetic Assessment
| Autor: | Albert Bruns, Susanne Boll, Benjamin Meyer, Larbi Abdenebaoui |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Exploit
Artificial neural network Computer science business.industry Deep learning 020206 networking & telecommunications 02 engineering and technology Machine learning computer.software_genre Convolutional neural network Field (computer science) 0202 electrical engineering electronic engineering information engineering Benchmark (computing) Task analysis 020201 artificial intelligence & image processing Artificial intelligence Adaptation (computer science) business computer |
| Zdroj: | CBMI |
| DOI: | 10.1109/cbmi.2018.8516273 |
| Popis: | Automatic photo assessment is a high emerging research field with wide useful ‘real-world’ applications. Due to the recent advances in deep learning, one can observe very promising approaches in the last years. However, the proposed solutions are adapted and optimized for ‘isolated’ datasets making it hard to understand the relationship between them and to benefit from the complementary information. Following a unifying approach, we propose in this paper a learning model that integrates the knowledge from different datasets. We conduct a study based on three representative benchmark datasets for photo assessment. Instead of developing for each dataset a specific model, we design and adapt sequentially a unique model which we nominate UNNA. UNNA consists of a deep convolutional neural network, that predicts for a given image three kinds of aesthetic information: technical quality, high-level semantical quality, and a detailed description of photographic rules. Due to the sequential adaptation that exploits the common features between the chosen datasets, UNNA has comparable performances with the state-of-the-art solutions with effectively less parameter. The final architecture of UNNA gives us some interesting indication of the kind of shared features as well as individual aspects of the considered datasets. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|