Workshop on Visual Concepts

In conjunction with The European Conference on Computer Vision (ECCV) 2024

Date: Sep 30 Afternoon | Location: Suite 4, MiCo Milano, Milan, Italy (Hybrid)

Concept learning and reasoning aim to extract compact and structured representations of the physical world, including sensing and actuation, and recompose them to tackle novel intricate problems. This paradigm suggests a unified view and solution to different recognition and generative tasks across different domains and modalities. An important research question is to understand and design concept representations that facilitate better learning from various datasets and compositional reasoning for novel tasks. As an endeavor to answering this question, in this workshop, we gather together researchers in computer vision, robotics, multi-modal learning, machine learning, and cognitive science to discuss the following topics:

Representations and computational models of concepts, such as object properties, relations, and actions;
Objectives and sources of learning signals to facilitate concept learning, incorporating insights from various scientific fields including machine learning, computer vision, natural language processing, robotics, and cognitive science;
Interpretability of existing learning systems, delving deeper into how these systems learn, represent, and make use of learned concepts in various application domains;
Applications of concept learning and reasoning, including but not limited to visual scene understanding, controllable visual generation, and robotics.

About

Concept learning and reasoning aim to extract compact and structured representations of the physical world, including sensing and actuation, and recompose them to tackle novel intricate problems. This paradigm suggests a unified view and solution to different recognition and generative tasks across different domains and modalities. An important research question is to understand and design concept representations that facilitate better learning from various datasets and compositional reasoning for novel tasks. As an endeavor to answering this question, in this workshop, we gather together researchers in computer vision, robotics, multi-modal learning, machine learning, and cognitive science to discuss the following topics:

Representations and computational models of concepts, such as object properties, relations, and actions;
Objectives and sources of learning signals to facilitate concept learning, incorporating insights from various scientific fields including machine learning, computer vision, natural language processing, robotics, and cognitive science;
Interpretability of existing learning systems, delving deeper into how these systems learn, represent, and make use of learned concepts in various application domains;
Applications of concept learning and reasoning, including but not limited to visual scene understanding, controllable visual generation, and robotics.