Leveraging inexpensive and human intervention-based annotating methodologies, such as crowdsourcing and web crawling, often leads to datasets with noisy labels. Noisy labels can have a detrimental impact on the performance and generalization of deep neural networks. Robust models that are able to handle and mitigate the effect of these noisy labels are thus essential. In this work, we explore the open challenges of neural network memorization and uncertainty in creating robust learning algorithms with noisy labels. To overcome them, we propose a novel framework called "Bayesian DivideMix++" with two critical components: (i) DivideMix++, to enhance the robustness against memorization and (ii) Monte-Carlo MixMatch, which focuses on improving the effectiveness towards label uncertainty. DivideMix++ improves the pipeline by integrating the warm-up and augmentation pipeline with self-supervised pre-training and dedicated different data augmentations for loss analysis and backpropagation. Monte-Carlo MixMatch leverages uncertainty measurements to mitigate the influence of uncertain samples by reducing their weight in the data augmentation MixMatch step. We validate our proposed pipeline using four datasets encompassing various synthetic and real-world noise settings. We demonstrate the effectiveness and merits of our proposed pipeline using extensive experiments. Bayesian DivideMix++ outperforms the state-of-the-art models by considerable differences in all experiments. Our findings underscore the potential of leveraging these modifications to enhance the performance and generalization of deep neural networks in practical scenarios.
1 July 2013 in Computer Methods in Biomechanics and Biomedical Engineering
Breast cancer screening is currently performed by mammography, which is limited by overlying anatomy and dense breast tissue. Computer-aided detection (CAD) systems can serve as a double reader to improve radiologist performance. In this paper, we have applied a novel approach to segmentation of suspicious region by mammogram and classification based on hybrid features with learning classifier. We formulated differentiation of lesion from normal tissue as a supervised learning problem, and applied this learning method to develop the classification algorithm. The algorithm has been verified with 164 mammograms in the mini Mammographic Image Analysis Society database. The experimental results show that the detection method has a sensitivity of 94.5% at 0.26 false positives per image. The efficiency of algorithm is measured using free receiver operating characteristics curve and the results are highlighted. We conclude that CAD technology with learning classifier has the potential to help radiologists with the task of discriminating between lesion and normal tissues.
Automatic recognition of basic emotions has improved the understanding of human emotions and has paved the path to understand complex emotions better, closer to real-world scenarios and having immediate applications. This paper investigates the utility of spectrogram (obtained from audio signals) images to perform complex emotion recognition by using transfer learning technique. Further, visual domain frameworks are adapted to audio frameworks. Transfer learning using pre-trained AlexNet are investigated. Information extracted from deep layers of the transfer learned networks was encoded as features and were trained on a linear classifier. Experimental results on EmoReact dataset consisting of 8 complex emotions shows the effectiveness of the proposed framework. Highest F1 score and AUC-ROC values as against the baseline performance by 16% and 10% respectively have been observed.
In the field of fine-grained food recognition, subset learning-based methods offer a strategic approach that groups classes into subsets to guide the training process. Our study introduces a novel approach, referred to as the Dining on Details (DoD), an innovative expert learning framework for food classification. This method ingeniously harnesses the power of large language models to construct subsets of classes within the dataset. The Dining on Details's efficacy is rooted in the robustness of the ImageBind multi-modality embedding space, which can identify meaningful similarities across varied categories. Trained through an end-to-end multi-task learning process, this method enhances performance in the fine-grained food recognition task, showing exceptional prowess with highly similar classes. A key advantage of DoD is its universal compatibility, allowing it to be applied seamlessly to any existing classification architecture. Our comprehensive validation of this method on various food datasets and backbones, both convolutional and transformer-based, reveals competitive results with significant performance gains ranging from 0.5% to 1.61%. Notably, it achieves state-of-the-art results on the Food-101 dataset.