Chapter
End-to-End Learning and Its Limitations in Machine Learning
59:08 - 1:07:31 (08:23)
The concept of end-to-end learning is often limited to end-to-end supervised learning for a specific task, which is a restricted view of the learning process. This approach divides vision into different modules for lower, mid and high levels of vision without considering the nuances of learning.
Clips
The perception of the external world involves creating entities, objects or structures, while perceptual organization deals with the process of defining these entities.
59:08 - 1:00:50 (01:42)
Summary
The perception of the external world involves creating entities, objects or structures, while perceptual organization deals with the process of defining these entities. The interaction between reorganization, recognition, and reconstruction is crucial in this process, with segmentation playing a small part in defining these entities.
ChapterEnd-to-End Learning and Its Limitations in Machine Learning
Episode#110 – Jitendra Malik: Computer Vision
PodcastLex Fridman Podcast
In computer vision, end-to-end learning is currently limited to supervised learning for a specific task, leading to a narrow view of the learning process.
1:00:51 - 1:05:37 (04:46)
Summary
In computer vision, end-to-end learning is currently limited to supervised learning for a specific task, leading to a narrow view of the learning process. This approach divides vision into low-level, mid-level, and high-level modules, which corresponds to the psychologist's notion of sensation, perception, and cognition.
ChapterEnd-to-End Learning and Its Limitations in Machine Learning
Episode#110 – Jitendra Malik: Computer Vision
PodcastLex Fridman Podcast
Children can learn about objects in the 3D world purely from visual signals, without needing to know the name of the object.
1:05:37 - 1:07:31 (01:53)
Summary
Children can learn about objects in the 3D world purely from visual signals, without needing to know the name of the object.