Abstract:Large-scale image datasets are often partially labeled, where only a few categories' labels are known for each image. Assigning pseudo-labels to unknown labels to gain additional training signals has become prevalent for training deep classification models. However, some pseudo-labels are inevitably incorrect, leading to a notable decline in the model classification performance. In this paper, we propose a novel method called Category-wise Fine-Tuning (CFT), aiming to reduce model inaccuracies caused by the wrong pseudo-labels. In particular, CFT employs known labels without pseudo-labels to fine-tune the logistic regressions of trained models individually to calibrate each category's model predictions. Genetic Algorithm, seldom used for training deep models, is also utilized in CFT to maximize the classification performance directly. CFT is applied to well-trained models, unlike most existing methods that train models from scratch. Hence, CFT is general and compatible with models trained with different methods and schemes, as demonstrated through extensive experiments. CFT requires only a few seconds for each category for calibration with consumer-grade GPUs. We achieve state-of-the-art results on three benchmarking datasets, including the CheXpert chest X-ray competition dataset (ensemble mAUC 93.33%, single model 91.82%), partially labeled MS-COCO (average mAP 83.69%), and Open Image V3 (mAP 85.31%), outperforming the previous bests by 0.28%, 2.21%, 2.50%, and 0.91%, respectively. The single model on CheXpert has been officially evaluated by the competition server, endorsing the correctness of the result. The outstanding results and generalizability indicate that CFT could be substantial and prevalent for classification model development. Code is available at:

Sample category label correction method and apparatus, and electronic device

Sample category label correction method and device, and electronic equipment

IMAGE PROCESSING METHOD AND DEVICE, ELECTRONIC APPARATUS AND STORAGE MEDIUM

Topic model based medical record classification method

The Correction Method Based on Optimal Solution in Digital Imaging system

OT Cleaner: Label Correction As Optimal Transport

HSALC: Hard Sample Aware Label Correction for Medical Image Classification

Image classification method and device and computer readable medium

Interactive Correction of Mislabeled Training Data

Correcting Sample Selection Bias for Image Classification

Medical image labeling method, system and device

Annotation data acquisition method and device, electronic equipment and storage medium

Multi-task meta label correction for time series prediction

An Adaptive Noisy Label-Correction Method Based on Selective Loss for Hyperspectral Image-Classification Problem

Confusing Pair Correction Based on Category Prototype for Domain Adaptation under Noisy Environments

Error-Bounded Correction of Noisy Labels

Learning From How Humans Correct

Co-Correcting: Noise-Tolerant Medical Image Classification via Mutual Label Correction

Learning to Purify Noisy Labels Via Meta Soft Label Corrector.

Category-wise Fine-Tuning: Resisting Incorrect Pseudo-Labels in Multi-Label Image Classification with Partial Labels

Reliable Label Correction is a Good Booster When Learning with Extremely Noisy Labels.