Researchers have developed a new method to improve visual recognition in datasets with noisy and long-tailed labels. The approach utilizes auxiliary text information from labels, leveraging pre-trained visual-language models to correct inconsistencies between images and their associated labels. This technique, termed Weak Teacher Supervision (WTS), is robust to label noise and data biases, showing superior performance on both synthetic and real-world datasets, especially under high-noise conditions. The source code for this method has been made publicly available. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Enhances robustness of visual recognition models in real-world, noisy datasets.
RANK_REASON Academic paper on a novel method for improving visual recognition with noisy labels.