Curricular SincNet: Towards Robust Deep Speaker Recognition by Emphasizing Hard Samples in Latent Space

Siamese network-based visual tracking algorithms have achieved excellent performance in recent years, but challenges such as fast target motion, shape and scale variations have made the tracking extremely difficult. The regression of anchor-free tracking has low computational complexity, strong real-time performance, and is suitable for visual tracking. Based on the anchor-free siamese tracking framework, this paper firstly introduces balance factors and modulation coefficients into the cross-entropy loss function to solve the classification inaccuracy caused by the imbalance between positive and negative samples as well as the imbalance between hard and easy samples during the training process, so that the model focuses more on the positive samples and the hard samples that make the major contribution to the training. Secondly, the intersection over union (IoU) loss function of the regression branch is improved, not only focusing on the IoU between the predicted box and the ground truth box, but also considering the aspect ratios of the two boxes and the minimum bounding box area that accommodate the two, which guides the generation of more accurate regression offsets. The overall loss of classification and regression is iteratively minimized and improves the accuracy and robustness of visual tracking. Experiments on four public datasets, OTB2015, VOT2016, UAV123 and GOT-10k, show that the proposed algorithm achieves the state-of-the-art performance.

Dynamic sampling for deep metric learning

Deep face recognition has achieved rapid development but still suffers from occlusions, illumination and pose variations, especially for face identification. The success of deep learning models in face recognition lies in large-scale high quality face data with accurate labels. However, in real-world applications, the collected data may be mixed with severe label noise, which significantly degrades the generalization ability of deep models. To alleviate the impact of label noise on face recognition, inspired by curriculum learning, we propose a self-paced deep learning model (SPDL) by introducing a negative \(l_1\)-norm regularizer for face recognition with label noise. During training, SPDL automatically evaluates the cleanness of samples in each batch and chooses cleaner samples for training while abandons the noisy samples. To demonstrate the effectiveness of SPDL, we use deep convolution neural network architectures for the task of robust face recognition. Experimental results show that our SPDL achieves superior performance on LFW, MegaFace and YTF when there are different levels of label noise.

Face recognition has made extraordinary progress owing to the advancement of deep convolutional neural networks (CNNs). The central task of face recognition, including face verification and identification, involves face feature discrimination. However, the traditional softmax loss of deep CNNs usually lacks the power of discrimination. To address this problem, recently several loss functions such as center loss, large margin softmax loss, and angular softmax loss have been proposed. All these improved losses share the same idea: maximizing inter-class variance and minimizing intra-class variance. In this paper, we propose a novel loss function, namely large margin cosine loss (LMCL), to realize this idea from a different perspective. More specifically, we reformulate the softmax loss as a cosine loss by L2 normalizing both features and weight vectors to remove radial variations, based on which a cosine margin term is introduced to further maximize the decision margin in the angular space. As a result, minimum intra-class variance and maximum inter-class variance are achieved by virtue of normalization and cosine decision margin maximization. We refer to our model trained with LMCL as CosFace. Extensive experimental evaluations are conducted on the most popular public-domain face recognition datasets such as MegaFace Challenge, Youtube Faces (YTF) and Labeled Face in the Wild (LFW). We achieve the state-of-the-art performance on these benchmarks, which confirms the effectiveness of our proposed approach.

In the biometrics community, face and speaker recognition are mature fields in which several systems have been proposed over the past twenty years. While existing systems perform well under controlled recording conditions, mismatch caused by the use of different sensors or a lack of cooperation from the subject still significantly affects performance, especially in challenging scenarios such as in forensics. Furthermore, existing methods suffer from scalability issues, which prevents them from taking advantage of increasingly large amounts of training data. This is otherwise a promising approach to improve accuracy in such challenging scenarios. In this thesis we address these problems of mismatch and complexity by developing scalable probabilistic models that we apply to face, speaker and bimodal recognition. Our contributions are four-fold. First, we propose a unified framework for session variability modeling techniques based on Gaussian mixture models (GMM), that encompasses inter-session variability (ISV) modeling, joint factor analysis (JFA) and total variability (TV) modeling. Second, we propose a novel exact and scalable formulation of probabilistic linear discriminant analysis (PLDA), which is a probabilistic and generative framework that models between-class and within-class variations. This formulation solves a major scalability issue, by improving both the time complexity of the training procedure from cubic to linear with respect to the number of samples per class, and the complexity of the scoring procedure. Furthermore, the implementations of all the proposed techniques are integrated into a novel collaborative open source software library called Bob 1 that enforces fair evaluations and encourages reproducible research. Fourth and finally, large-scale experiments are conducted with all of the above machine learning algorithms on several databases such as FRGC for face recognition, NIST SRE12 for speaker recognition and MOBIO for bimodal recognition, showing competitive performances.

Deep CNN models have become state-of-the-art techniques in many application, e.g., face recognition, speaker recognition, and image classification. Although many studies address on speedup or compression of individual models, very few studies focus on co-compressing and unifying models from different modalities. In this work, to joint and compress face and speaker recognition models, a shared-codebook approach is adopted to reduce the redundancy of the combined model. Despite the modality of the inputs of these two CNN models are quite different, the shared codebook can support two CNN models of sound and image for speaker and face recognition. Experiments show the promising results of unified and co-compressing heterogeneous models for efficient inference.

Current training objectives of existing person Re-IDentification (ReID) models only ensure that the loss of the model decreases on selected training batch, with no regards to the performance on samples outside the batch. It will inevitably cause the model to over-fit the data in the dominant position (e.g., head data in imbalanced class, easy samples or noisy samples). The latest resampling methods address the issue by designing specific criterion to select specific samples that trains the model generalize more on certain type of data (e.g., hard samples, tail data), which is not adaptive to the inconsistent real world ReID data distributions. Therefore, instead of simply presuming on what samples are generalizable, this paper proposes a one-for-more training objective that directly takes the generalization ability of selected samples as a loss function and learn a sampler to automatically select generalizable samples. More importantly, our proposed one-for-more based sampler can be seamlessly integrated into the ReID training framework which is able to simultaneously train ReID models and the sampler in an end-to-end fashion. The experimental results show that our method can effectively improve the ReID model training and boost the performance of ReID models.

Knowledge distillation is a representative technique for model compression and acceleration, which is important for deploying neural networks on resource limited devices. The knowledge transferred from teacher to student is the mapping of teacher model, or represented by all the input-output pairs. However, in practice the student model only learns from data pairs of the dataset that may be biased, and we think this limits the performance of knowledge distillation. In this paper, we first quantitatively define the uniformity of the sampled data for training, providing a unified view for methods that learn from biased data. Then we evaluate the uniformity on real world dataset and show that existing methods actually improve the uniformity of data. We further introduce two uniformity-oriented methods for rectifying the bias of data for knowledge distillation. Extensive experiments conducted on Face Recognition and Person Re-identification have shown the effectiveness of our method. Moreover, we analyze the sampled data on Face Recognition and show that better balance is achieved between races and between easy and hard samples. And this effect can be also confirmed in training the student model from scratch, resulting in a comparable performance with standard knowledge distillation.

Breast ultrasound image classification with hard sample generation and semi-supervised learning

Convolutional neural networks (CNNs) have shown great advantages in computer vision fields, and loss functions are of great significance to their gradient descent algorithms. Softmax loss, a combination of cross-entropy loss and Softmax function, is the most commonly used one for CNNs. Hence, it can continuously increase the discernibility of sample features in classification tasks. Intuitively, to promote the discrimination of CNNs, the learned features are desirable when the inter-class separability and intra-class compactness are maximized simultaneously. Since Softmax loss hardly motivates this inter-class separability and intra-class compactness simultaneously and explicitly, we propose a new method to achieve this simultaneous maximization. This method minimizes the distance between features of homogeneous samples along with Softmax loss and thus improves CNNs' performance on vision-related tasks. Experiments on both visual classification and face verification datasets validate the effectiveness and advantages of our method.

Metaheuristics based COVID-19 detection using medical images: A review

Federated Learning (FL) enables massive local data owners to collaboratively train a deep learning model without disclosing their private data. The importance of local data samples from various data owners to FL models varies widely. This is exacerbated by the presence of noisy data that exhibit large losses similar to important (hard) samples. Currently, there lacks an FL approach that can effectively distinguish hard samples (which are beneficial) from noisy samples (which are harmful). To bridge this gap, we propose the joint Federated Meta-Weighting based Client and Sample Selection (FedMW-CSS) approach to simultaneously mitigate label noise and hard sample selection. It is a bilevel optimization approach for FL client-and-sample selection and global model construction to achieve hard sample-aware noise-robust learning in a privacy preserving manner. It performs meta-learning based online approximation to iteratively update global FL models, select the most positively influential samples and deal with training data noise. To utilize both the instance-level information and class-level information for better performance improvements, FedMW-CSS efficiently learns a class-level weight by manipulating gradients at the class level, e.g., it performs a gradient descent step on class-level weights, which only relies on intermediate gradients. Theoretically, we analyze the privacy guarantees and convergence of FedMW-CSS. Extensive experiments comparison against eight state-of-the-art baselines on six real-world datasets in the presence of data noise and heterogeneity shows that FedMW-CSS achieves up to 28.5% higher test accuracy, while saving communication and computation costs by at least 49.3% and 1.2%, respectively.

Face recognition has witnessed significant progress due to the advances of deep convolutional neural networks (CNNs), the central task of which is how to improve the feature discrimination. To this end, several margin-based (e.g., angular, additive and additive angular margins) softmax loss functions have been proposed to increase the feature margin between different classes. However, despite great achievements have been made, they mainly suffer from three issues: 1) Obviously, they ignore the importance of informative features mining for discriminative learning; 2) They encourage the feature margin only from the ground truth class, without realizing the discriminability from other non-ground truth classes; 3) The feature margin between different classes is set to be same and fixed, which may not adapt the situations very well. To cope with these issues, this paper develops a novel loss function, which adaptively emphasizes the mis-classified feature vectors to guide the discriminative feature learning. Thus we can address all the above issues and achieve more discriminative face features. To the best of our knowledge, this is the first attempt to inherit the advantages of feature margin and feature mining into a unified loss function. Experimental results on several benchmarks have demonstrated the effectiveness of our method over state-of-the-art alternatives. Our code is available at http://www.cbsr.ia.ac.cn/users/xiaobowang/.

For communication, speech is one of the natural forms. A person’s voice contains various parameters that convey information such as emotions, gender, attitude, health, and identity. Determination of these parameters will help to further develop the technology into a reliable and consistent means of identification using speaker recognition. Speaker recognition technologies have wide application areas especially in authentication; surveillance and forensic speaker recognition. In addition, speaker recognition refers to the automated method of identifying or confirming the identity of an individual based on his/her voice. Speech recognition strips out the personal differences to detect the words. Speaker recognition typically disregards the language and meaning to detect the physical person behind the speech. Speech recognition is language-dependent, while Speaker recognition is independent of language. In essence, voice biometrics provides speaker recognition rather than speech recognition. The most accepted form of identification for a human is his/her speech signal. Principally the speaker recognition is the computing task of validating a user’s claimed identity 108using characteristics extracted from their voice. The speaker recognition process based on a speech signal is treated as one of the most exciting technologies of human recognition. For Speaker identification activities we mainly emphasize the physical features of signal. Speakers could be categorized as speaker identification and speaker verification. In speaker identification, the obtained features are compared with all the speaker’s features which are stored in a voice model database and in speaker verification the obtained features are only compared with the stored features of the speaker he/she claimed to be. In this chapter the general principles of speaker recognition, methodology, and applications are discussed.

Survey on Deep Neural Networks in Speech and Vision Systems