Multilabel classification outperforms detection-based technique

Ronit Gross, Ella Koresh, Tal Halevi, Shiri Hodassman, Yuval Meir, Yarden Tzach and Ido Kanter

Physica A: Statistical Mechanics and its Applications, 2025, vol. 658, issue C

Abstract: In real-life scenarios, an input image typically comprises multiple objects, and their classification is often implemented using detection-based classification (DBC). In this approach, objects are first detected and then identified individually using a deep architecture. In this study, we demonstrate that the accuracy achieved by multilabel classification (MLC) surpasses that of DBC for a relatively small number of multilabel learning combinations. The crossover point at which DBC maximizes accuracy depends on the type of multilabel images, such as the number of multiple objects per image. The results are based on VGG-6 trained on the CIFAR-100 dataset using an upper bound for DBC accuracy, assumed under perfect detection conditions. Furthermore, we briefly discuss the potential relevance of these findings to advanced communication theory and natural language processing. The results suggest a need to reexamine the advantages of MLC over DBC using more complex datasets and deep architectures.

Keywords: Deep learning; Machine learning; Shallow learning (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0378437124008057
Full text for ScienceDirect subscribers only. Journal offers the option of making the article available online on Science direct for a fee of $3,000

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:phsmap:v:658:y:2025:i:c:s0378437124008057

DOI: 10.1016/j.physa.2024.130295

Access Statistics for this article

Physica A: Statistical Mechanics and its Applications is currently edited by K. A. Dawson, J. O. Indekeu, H.E. Stanley and C. Tsallis

More articles in Physica A: Statistical Mechanics and its Applications from Elsevier
Bibliographic data for series maintained by Catherine Liu ().