The course is divided into four sections, or modules, describing the different stages of visual intelligence:
- Module 1: Visual Stream. It explores experimental and theoretical aspects of hierarchical models of the ventral stream -- from HMAX to recent deep learning models.
- Module 2: Memory and Executive Function. It explores a set of routines to integrate vision and cognition. This includes routines critical for understanding a scene, such as planning the next saccade and incorporating the visual information into working memory, as well as planning navigation and grasping.
- Module 3: The Cognitive Core. This module explores the algorithms and circuit hardware involved in acquiring world knowledge that go beyond what is in the image and using it for visual understanding.
- Module 4: Symbolic Compositional Models. It explores higher-level symbolic processes that are important for human visual understanding, including inferences on images and videos.
Each module will be introduced and discussed through 2 or 3 lectures that depict current theoretical aspects and experimental evidence. Each speaker will start and conclude the talk by situating their module in the context of the other ones.
- outline the present research plan for the module for the next 5 years, stressing what is already cast in stone and what is still open
- provide the necessary background for the students
- discuss -- and stimulate participants to think about -- possible interactions between modules.
- challenge students to contribute ideas and to improve our plans
The class is an interesting opportunity for developing further our preliminary plans of the CBMM research proposal. For this reason, we encourage CBMM members to attend -- and contribute to -- the course.