2 weeks ago
1
A probe insubstantial from the Indian Institute of Science (IISc) has been selected arsenic the champion insubstantial finalist astatine the IEEE/CVF Conference connected Computer Vision and Pattern Recognition (CVPR) 2026.
The CVPR is 1 of the world’s premier conferences successful machine imaginativeness and artificial quality (AI), and this year’s league was held astatine Colorado successful the U.S.
The IISc team, comprising R. Venkatesh Babu, Priyam Dey, Aditya Sahdev, Sunny Bhati, Konda Reddy Mopuri from the Department of Computational and Data Sciences (CDS), was selected arsenic the champion insubstantial finalists for the insubstantial connected “Rethinking Dataset Distillation: Hard Truths astir Soft Labels”.
Prof. Babu, who is the seat of the CDS, said that the insubstantial focuses connected the precise important taxable of dataset distillation.
“This (paper) is astir dataset distillation. There is simply a immense magnitude of data, and we usage it for grooming assorted models. We don’t cognize whether each the information we usage for grooming are utile oregon not. We don’t request millions of samples, lone thousands of samples are capable to bid the model,” Prof. Babu told The Hindu.
He added that isolated from helping successful information businesslike learning and improving scaling show of the models, their probe besides helps bring down the c emission.
“Dataset distillation has emerged arsenic a promising absorption for businesslike learning, and this enactment revisits immoderate of its halfway assumptions, uncovering important insights astir the relation of brushed labels and their limitations,” the CDS said.
The IISc insubstantial was recognised among the apical 15 papers retired of much than 16,000 submissions made astatine the CVPR 2026.
“From 16,000 submissions, astir 4,000 papers were selected and from this we made it to the apical 15, which is simply a important accomplishment for the team,” Prof. Babu added.
