Unsupervised part representation by Flow Capsules

by   Sara Sabour, et al.

Capsule networks are designed to parse an image into a hierarchy of objects, parts and relations. While promising, they remain limited by an inability to learn effective low level part descriptions. To address this issue we propose a novel self-supervised method for learning part descriptors of an image. During training, we exploit motion as a powerful perceptual cue for part definition, using an expressive decoder for part generation and layered image formation with occlusion. Experiments demonstrate robust part discovery in the presence of multiple objects, cluttered backgrounds, and significant occlusion. The resulting part descriptors, a.k.a. part capsules, are decoded into shape masks, filling in occluded pixels, along with relative depth on single images. We also report unsupervised object classification using our capsule parts in a stacked capsule autoencoder.


page 6

page 7


DPR-CAE: Capsule Autoencoder with Dynamic Part Representation for Image Parsing

Parsing an image into a hierarchy of objects, parts, and relations is im...

HP-Capsule: Unsupervised Face Part Discovery by Hierarchical Parsing Capsule Network

Capsule networks are designed to present the objects by a set of parts a...

SS-3DCapsNet: Self-supervised 3D Capsule Networks for Medical Segmentation on Less Labeled Data

Capsule network is a recent new deep network architecture that has been ...

Stacked Capsule Autoencoders

An object can be seen as a geometrically organized set of interrelated p...

Canonical Capsules: Unsupervised Capsules in Canonical Pose

We propose an unsupervised capsule architecture for 3D point clouds. We ...

Unsupervised Layered Image Decomposition into Object Prototypes

We present an unsupervised learning framework for decomposing images int...

Unsupervised Discovery of Parts, Structure, and Dynamics

Humans easily recognize object parts and their hierarchical structure by...