NIPS Workshops, Whistler 2009

Scalable Learning in Computer Vision

author: Adam Coates, Department of Computer Science, Stanford University
published: Jan. 19, 2010, recorded: December 2009, views: 444

Slides

Slides
0:00	Scalable Learning in Computer Vision
0:33	Computer Vision is Hard
1:02	Introduction -1
1:37	Introduction -2
2:25	Introduction -3
3:11	Supervised learning
3:19	The Learning Pipeline -1
3:58	Synthetic Data -1
4:55	Synthetic Data -2
5:25	Synthetic Data: Example
6:00	The Learning Pipeline -2
6:31	Features on CPUs vs. GPUs
7:04	Features on GPUs
8:08	The Learning Pipeline -3
9:00	Distributed Training -1
9:33	Distributed Training -2
10:20	The Learning Pipeline -4
11:06	Size Matters
11:43	Unsupervised Feature Learning -1
12:10	Traditional supervised learning
13:03	Self-taught learning
13:25	Learning representations
13:37	Computer vision features
14:09	Unsupervised Feature Learning -2
14:55	Unsupervised Feature Learning -3
15:52	Learning Large RBMs on GPUs
16:44	Learning features
17:53	Conclusion
19:38	Thank you

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.

Description

Computer vision is a challenging application area of machine learning. Recent work has shown that large training sets may yield higher performance in vision tasks like object detection. We overview our work in object detection using a scalable, distributed training system capable of training on more than 100 million examples in just a few hours. We also briefly describe recent work with deep learning algorithms that may allow us to apply these architectures to large datasets as well.