Attendees are seated in a conference room listening to a presentation on image classification at the CVPR (Conference on Computer Vision and Pattern Recognition). The slide displayed on the screen is titled "Example: Image Classification" and presents a specific study, "SeLA (Asano et al., 2020)", published by the Australian National University. The slide illustrates a conceptual diagram showing the image classification process involving inputs, convolution (Convnet), backpropagation, and classification with pseudo-labels. The focus of the idea presented is to use "optimal transport (OT)" for label generation. The detailed reference to the work by Ming Xu and Stephen Gould, "Temporally Consistent Unbalanced Optimal Transport for Unsupervised Image Classification," is visible at the bottom of the slide. The conference setting includes rows of attendees, some of whom are observing, taking notes, or thoughtfully engaging with the presented material. The industrial-style ceiling with exposed ductwork provides a modern and professional ambience to the event.
Text transcribed from the image:
Example: Image Classification
Example: SeLA (Asano et al., 2020)
Figure courtesy Caron et al, 2018
Input
Convnet
Backprop
Classification
Pseudo-labels
Optimal
Transport
(OT)
> Idea: Use optimal transport (OT) for label generation!
Ming Xu and Stephen Gould. Temporally Consistent Unbalanced Optimal Transport for Unsupervised
eg
Australian
National
University
יחות
AD