Corpus ID: 63828210

RotationNet: Joint Learning of Object Classification and Viewpoint Estimation using Unaligned 3D Object Dataset

@inproceedings{Kanezaki2016RotationNetJL,
  title={RotationNet: Joint Learning of Object Classification and Viewpoint Estimation using Unaligned 3D Object Dataset},
  author={Asako Kanezaki and Y. Matsushita and Y. Nishida},
  year={2016}
}

• Asako Kanezaki, Y. Matsushita, Y. Nishida
• Published 2016
• Computer Science

We propose a Convolutional Neural Network (CNN)based model “RotationNet,” which takes multi-view images of an object as input and estimates both its pose and object category. Unlike previous approaches that use known pose labels for training, our method treats the pose labels as latent variables, which are optimized to self-align in an unsupervised manner during the training using an unaligned dataset. RotationNet is designed to use only a partial set of multi-view images for inference, and… Expand

View PDF on arXiv

Save to Library

Create Alert

Cite

Launch Research Feed