Using Early-Learning Regularization to Classify Real-World Noisy Data
This work addresses label noise in computer vision for real-world applications, but it is incremental as it replicates and extends prior methods.
The study tackled the problem of classifying noisy real-world data by replicating Early-Learning Regularization and testing it on a real-world dataset, achieving consistent results and a 14.6 percentage point improvement with Sharpness-Aware Minimization.
The memorization problem is well-known in the field of computer vision. Liu et al. propose a technique called Early-Learning Regularization, which improves accuracy on the CIFAR datasets when label noise is present. This project replicates their experiments and investigates the performance on a real-world dataset with intrinsic noise. Results show that their experimental results are consistent. We also explore Sharpness-Aware Minimization in addition to SGD and observed a further 14.6 percentage points improvement. Future work includes using all 6 million images and manually clean a fraction of the images to fine-tune a transfer learning model. Last but not the least, having access to clean data for testing would also improve the measurement of accuracy.