Seeing the Intangible: Survey of Image Classification into High-Level and Abstract Categories

2023-08-21 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

A survey paper by Delfina Sol Martinez Pandiani and Valentina Presutti, submitted on August 21, 2023, and last revised on February 29, 2024, systematically reviews research on high-level visual understanding in Computer Vision (CV), specifically focusing on Abstract Concepts (ACs) in automatic image classification. The authors clarify the tacit understanding of high-level semantics in CV through a multidisciplinary analysis, categorizing them into clusters like commonsense, emotional, aesthetic, and inductive interpretative semantics. The survey identifies and categorizes CV tasks associated with high-level visual sensemaking and examines how abstract concepts such as values and ideologies are handled in CV. It highlights persistent challenges, including the limited efficacy of massive datasets and the importance of integrating supplementary information and mid-level features, emphasizing the growing relevance of hybrid AI systems.

Key takeaway

For research scientists developing advanced Computer Vision systems, understanding the nuances of abstract concept classification is critical. You should prioritize integrating supplementary information and mid-level features into your models, as large datasets alone are proving insufficient. Consider exploring hybrid AI architectures to address the multifaceted nature of high-level visual reasoning tasks.

Key insights

High-level visual understanding in CV requires clarifying abstract concepts and integrating diverse data sources.

Principles

• High-level semantics are multidisciplinary.
• Massive datasets alone are insufficient for ACs.

In practice

• Integrate supplementary information.
• Utilize mid-level features.
• Explore hybrid AI systems.

Topics

• Computer Vision
• High-Level Visual Understanding
• Abstract Concepts
• Image Classification
• Visual Sensemaking

Best for: Research Scientist, AI Scientist, Computer Vision Engineer

Related on AIssential

• Deep Model for Vision — Deep learning overcomes classical computer vision limitations, enabling scalable and adaptable solutions across diverse visual tasks.
• Human-level 3D shape perception emerges from multi-view learning — Human-level 3D perception emerges from multi-view learning using a visual-spatial objective on naturalistic data.
• Learning to see the physical world: an interview with Jiajun Wu — Physical scene understanding requires integrating diverse models and leveraging structural information for data-efficient learning.
• Characterizing the visual representation of objects from the child's view — Children's visual object experiences are skewed and variable, yet reveal strong superordinate category structure.
• Are AI vision tasks still underused? — Specialized vision models are underutilized due to a general lack of awareness regarding their specific capabilities.
• Seeing Together:Multi-Robot Cooperative Egocentric Spatial Reasoning with Multimodal Large Language Models — Multi-robot egocentric spatial reasoning benefits from physics-informed MLLMs and specialized cooperative datasets.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.