After taking this course students will be able to:
(Recognition)
-
Identify potential applications of recognition
-
Identify who would benefit from each
application, who might be harmed, what the input data might look like, and the
quality and nature of output needed.
-
Discuss the pros and cons of particular
recognition tasks, benchmarks and metrics in the context of possible applications,
and potentially define new ones.
-
Understand and describe the dominant technical
approaches to various recognition problems : image
classification, object detection, segmentation and pose estimation
-
Identify the current research challenges and
their impact on the actual applications.
(Reconstruction)
-
Identify potential applications of
reconstruction, and describe benefits, harms, nature of input data and quality
and nature of needed output
-
Derive equations describing image formation and various
invariants thereof (e.g., epipolar constraint)
-
Describe the classical approaches based on
estimating correspondence and define the key challenges involved.
-
Describe modern end-to-end approaches and their
benefits and limitations.
-
Discuss the pros and cons of various
representations of 3D shape
(Embodied cognition)
-
Define what computer vision means for embodied
agent, and articulate how this is different from other applications previously
encountered
-
Describe research challenges in computer vision
for embodied agents, including those of learning, control and interaction with
humans
(How to do computer vision research)
-
Concretely identify a research question of
significance
-
Review prior work and crisply identify limitations
-
Design new solutions to address identified
limitations
-
Evaluate and audit the proposed solution with
the actual application and potential ethical considerations in mind.
-
Provide feedback and guidance to peer
researchers.
-
Write a technical paper that can be accepted to
a workshop or conference.