Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY

Abstract: 

Unsupervised object discovery in images involves uncovering recurring patterns that define objects and discriminates them against the background. This is more challenging than image clustering as the size and the location of the objects are not known: this adds additional degrees of freedom and increases the problem complexity. In this work, we propose StampNet, a novel autoencoding neural network that localizes shapes (objects) over a simple background in images and categorizes them simultaneously. StampNet consists of a discrete latent space that is used to categorize objects and to determine the location of the objects. The object categories are formed during the training, resulting in the discovery of a fixed set of objects. We present a set of experiments that demonstrate that StampNet is able to localize and cluster multiple overlapping shapes with varying complexity including the digits from the MNIST dataset. We also present an application of StampNet in the localization of pedestrians in overhead depth-maps.

up
0 users have voted:

Paper Details

Authors:
Joost Visser, Alessandro Corbetta, Vlado Menkovski, Federico Toschi
Submitted On:
20 September 2019 - 11:41am
Short Link:
Type:
Poster
Event:
Presenter's Name:
Vlado Menkovski
Paper Code:
2794
Document Year:
2019
Cite

Document Files

StampNet.pdf

(9)

Subscribe

[1] Joost Visser, Alessandro Corbetta, Vlado Menkovski, Federico Toschi, "STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4788. Accessed: Oct. 18, 2019.
@article{4788-19,
url = {http://sigport.org/4788},
author = {Joost Visser; Alessandro Corbetta; Vlado Menkovski; Federico Toschi },
publisher = {IEEE SigPort},
title = {STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY},
year = {2019} }
TY - EJOUR
T1 - STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY
AU - Joost Visser; Alessandro Corbetta; Vlado Menkovski; Federico Toschi
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4788
ER -
Joost Visser, Alessandro Corbetta, Vlado Menkovski, Federico Toschi. (2019). STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY. IEEE SigPort. http://sigport.org/4788
Joost Visser, Alessandro Corbetta, Vlado Menkovski, Federico Toschi, 2019. STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY. Available at: http://sigport.org/4788.
Joost Visser, Alessandro Corbetta, Vlado Menkovski, Federico Toschi. (2019). "STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY." Web.
1. Joost Visser, Alessandro Corbetta, Vlado Menkovski, Federico Toschi. STAMPNET: UNSUPERVISED MULTI-CLASS OBJECT DISCOVERY [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4788