Loading...
Please wait, while we are loading the content...
Similar Documents
Visual Tasks beyond Categorization for Training Convolutional Neural Networks
| Content Provider | Semantic Scholar |
|---|---|
| Author | James Dicarlo |
| Copyright Year | 2016 |
| Abstract | Humans can perceive a variety of visual properties of objects besides their category. In this paper, we explorewhether convolutional neural networks (CNNs) can also learn object-related variables. The models are trained for object position, size and pose, respectively, from synthetic images and tested on unseen held-out objects. First, we show that some object properties come "for free" from learning others, and poseoptimized model can generalize to both categorical and non-categorical variables. Second, we demonstrate that pre-training the model with pose facilitates learning object categories from both synthetic and realistic images. Thesis Supervisor: James J. DiCarlo Title: Professor of Neuroscience; Head, Department of Brain and Cognitive Sciences |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://dspace.mit.edu/bitstream/handle/1721.1/106095/965383395-MIT.pdf?sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
