Semantic Jitter: Dense Supervision for Visual Comparisons via Synthetic Images

12/19/2016
by   Aron Yu, et al.
0

Distinguishing subtle differences in attributes is valuable, yet learning to make visual comparisons remains non-trivial. Not only is the number of possible comparisons quadratic in the number of training images, but also access to images adequately spanning the space of fine-grained visual differences is limited. We propose to overcome the sparsity of supervision problem via synthetically generated images. Building on a state-of-the-art image generation engine, we sample pairs of training images exhibiting slight modifications of individual attributes. Augmenting real training image pairs with these examples, we then train attribute ranking models to predict the relative strength of an attribute in novel pairs of real images. Our results on datasets of faces and fashion images show the great promise of bootstrapping imperfect image generators to counteract sample sparsity for learning to rank.

READ FULL TEXT

page 1

page 3

page 4

page 5

page 12

research
01/08/2019

Thinking Outside the Pool: Active Training Image Creation for Relative Attributes

Current wisdom suggests more labeled image data is always better, and ob...
research
03/31/2018

Compare and Contrast: Learning Prominent Visual Differences

Relative attribute models can compare images in terms of all detected pr...
research
09/09/2020

Relative Attribute Classification with Deep Rank SVM

Relative attributes indicate the strength of a particular attribute betw...
research
11/26/2021

TRIP: Refining Image-to-Image Translation via Rival Preferences

Relative attribute (RA), referring to the preference over two images on ...
research
12/13/2015

Deep Relative Attributes

Visual attributes are great means of describing images or scenes, in a w...
research
03/08/2015

Understanding Image Virality

Virality of online content on social networking websites is an important...
research
05/24/2021

Large-Scale Attribute-Object Compositions

We study the problem of learning how to predict attribute-object composi...

Please sign up or login with your details

Forgot password? Click here to reset