Unsupervised learning of clutter-resistant visual representations from natural videos
Populations of neurons in inferotemporal cortex (IT) maintain an explicit code for object identity that also tolerates transformations of object appearance e.g., position, scale, viewing angle [1, 2, 3]. Though the learning rules are not known, recent results [4, 5, 6] suggest the operation of an un...
| Main Authors: | , , |
|---|---|
| Format: | Technical Report |
| Language: | en_US |
| Published: |
Center for Brains, Minds and Machines (CBMM), arXiv
2015
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/1721.1/100187 |
| _version_ | 1826208915405144064 |
|---|---|
| author | Liao, Qianli Leibo, Joel Z Poggio, Tomaso |
| author_facet | Liao, Qianli Leibo, Joel Z Poggio, Tomaso |
| author_sort | Liao, Qianli |
| collection | MIT |
| description | Populations of neurons in inferotemporal cortex (IT) maintain an explicit code for object identity that also tolerates transformations of object appearance e.g., position, scale, viewing angle [1, 2, 3]. Though the learning rules are not known, recent results [4, 5, 6] suggest the operation of an unsupervised temporal-association-based method e.g., Foldiak’s trace rule [7]. Such methods exploit the temporal continuity of the visual world by assuming that visual experience over short timescales will tend to have invariant identity content. Thus, by associating representations of frames from nearby times, a representation that tolerates whatever transformations occurred in the video may be achieved. Many previous studies verified that such rules can work in simple situations without background clutter, but the presence of visual clutter has remained problematic for this approach. Here we show that temporal association based on large class-specific filters (templates) avoids the problem of clutter. Our system learns in an unsupervised way from natural videos gathered from the internet, and is able to perform a difficult unconstrained face recognition task on natural images (Labeled Faces in the Wild [8]). |
| first_indexed | 2024-09-23T14:14:38Z |
| format | Technical Report |
| id | mit-1721.1/100187 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T14:14:38Z |
| publishDate | 2015 |
| publisher | Center for Brains, Minds and Machines (CBMM), arXiv |
| record_format | dspace |
| spelling | mit-1721.1/1001872019-04-10T19:04:58Z Unsupervised learning of clutter-resistant visual representations from natural videos Liao, Qianli Leibo, Joel Z Poggio, Tomaso Object Recognition Computer vision Machine Learning Artificial Intelligence Populations of neurons in inferotemporal cortex (IT) maintain an explicit code for object identity that also tolerates transformations of object appearance e.g., position, scale, viewing angle [1, 2, 3]. Though the learning rules are not known, recent results [4, 5, 6] suggest the operation of an unsupervised temporal-association-based method e.g., Foldiak’s trace rule [7]. Such methods exploit the temporal continuity of the visual world by assuming that visual experience over short timescales will tend to have invariant identity content. Thus, by associating representations of frames from nearby times, a representation that tolerates whatever transformations occurred in the video may be achieved. Many previous studies verified that such rules can work in simple situations without background clutter, but the presence of visual clutter has remained problematic for this approach. Here we show that temporal association based on large class-specific filters (templates) avoids the problem of clutter. Our system learns in an unsupervised way from natural videos gathered from the internet, and is able to perform a difficult unconstrained face recognition task on natural images (Labeled Faces in the Wild [8]). This work was supported by the Center for Brains, Minds and Machines (CBMM), funded by NSF STC award CCF-1231216. 2015-12-10T23:55:38Z 2015-12-10T23:55:38Z 2015-04-27 Technical Report Working Paper Other http://hdl.handle.net/1721.1/100187 arXiv:1409.3879v2 en_US CBMM Memo Series;023 Attribution-NonCommercial 3.0 United States http://creativecommons.org/licenses/by-nc/3.0/us/ application/pdf Center for Brains, Minds and Machines (CBMM), arXiv |
| spellingShingle | Object Recognition Computer vision Machine Learning Artificial Intelligence Liao, Qianli Leibo, Joel Z Poggio, Tomaso Unsupervised learning of clutter-resistant visual representations from natural videos |
| title | Unsupervised learning of clutter-resistant visual representations from natural videos |
| title_full | Unsupervised learning of clutter-resistant visual representations from natural videos |
| title_fullStr | Unsupervised learning of clutter-resistant visual representations from natural videos |
| title_full_unstemmed | Unsupervised learning of clutter-resistant visual representations from natural videos |
| title_short | Unsupervised learning of clutter-resistant visual representations from natural videos |
| title_sort | unsupervised learning of clutter resistant visual representations from natural videos |
| topic | Object Recognition Computer vision Machine Learning Artificial Intelligence |
| url | http://hdl.handle.net/1721.1/100187 |
| work_keys_str_mv | AT liaoqianli unsupervisedlearningofclutterresistantvisualrepresentationsfromnaturalvideos AT leibojoelz unsupervisedlearningofclutterresistantvisualrepresentationsfromnaturalvideos AT poggiotomaso unsupervisedlearningofclutterresistantvisualrepresentationsfromnaturalvideos |