Unsupervised learning with sparse space-and-time autoencoders
This work addresses unsupervised learning for sparse data in computer vision and motion capture, but appears incremental as it applies known autoencoder methods to new dimensional contexts.
The paper tackled modeling sparse structures in 2D, 3D, and 4D space-time using spatially-sparse convolutional autoencoders, and evaluated their latent spaces on downstream tasks like handwriting recognition and segmentation, achieving unspecified results without concrete numbers.
We use spatially-sparse two, three and four dimensional convolutional autoencoder networks to model sparse structures in 2D space, 3D space, and 3+1=4 dimensional space-time. We evaluate the resulting latent spaces by testing their usefulness for downstream tasks. Applications are to handwriting recognition in 2D, segmentation for parts in 3D objects, segmentation for objects in 3D scenes, and body-part segmentation for 4D wire-frame models generated from motion capture data.