Softmax Loss Research Articles

A unified perspective of classification-based loss and distance-based loss for cross-view gait recognition

Gait can be used to recognize people in an uncooperative and noninvasive manner and it is hard to imitate or counterfeit, which makes it suitable for video surveillance. The current solutions for gait recognition are still not robust to handle the conditions when the view angles of the gallery and query are different. We improve the performance of cross-view gait recognition from the perspective of metric learning. Specifically, we propose to use angular softmax loss to impose an angular margin for extracting separable features. At the same time, we use triplet loss to make the extracted features more discriminative. Additionally, we add a batch-normalization layer after extracting gait features to effectively optimize two different losses. We evaluate our approach on two widely-used gait dataset: CASIA-B dataset and TUM GAID dataset. The experiment results show that our approach outperforms the prior state-of-the-art approaches, which shows the effectiveness of our approach.

Masked Face Recognition From Synthesis to Reality

As we have been seriously hit by the COVID-19 pandemic, wearing a facial mask is a crucial action that we can take for our protection. This paper reports a comprehensive study on the recognition of masked faces. By using facial landmarks, we synthesize the facial mask for each face in several benchmark databases with different challenging factors. The IJB-B and IJB-C databases are selected for evaluating the performance against the variation across pose, illumination and expression (PIE). The FG-Net database is selected for evaluating the performance across age. The SCface is chosen for evaluating the performance on low-resolution images. The MS-1MV2 is exploited as the base training set. We use the ResNet-100 as the feature embedding network connected to state-of-the-art loss functions designed for tackling face recognition. The loss functions considered include the Center Loss, the Marginal Loss, the Angular Softmax Loss, the Large Margin Cosine Loss and the Additive Angular Margin Loss. Both verification and identification are conducted in our evaluation. The performances for recognizing faces with and without the synthetic masks are all evaluated for a complete comparison. The network with the best loss function for recognizing synthetic masked faces is then assessed on a real masked face database, the cleaned RMFRD (c-RMFRD) dataset. Compared with a human user test on the c-RMFRD, the network trained on the synthetic masked faces outperforms human vision for a large gap. Our contributions are fourfold. The first is a comprehensive study for tackling masked face recognition by using state-of-the-art loss functions against various compounding factors. For comparison purpose, the second is another comprehensive study on the recognition of faces without masks by using the same loss functions against the same challenging factors. The third is the verification of the network trained on synthetic masked faces for tackling the real masked face recognition with performance better than human inspectors. The fourth is the highlight on the challenges of masked face recognition and the directions for future research. Our code, trained models and dataset are available via the project GitHub site.

A multi-terminal pushing method for emergency information in a smart city based on deep learning

In order to overcome the problems of large information identification error and high error rate in traditional emergency information push methods, a multi-terminal push method of emergency information in a smart city based on deep learning is proposed. The convolution neural network based on deep learning is used to identify the real-time collection results of intelligent city operation monitoring information, and the combination of softmax loss function and centre loss function is used to reduce the identification error. Through the process of retrieval, adaptation and detection of crisis cases, crisis types are determined and emergency decision-making is selected. Based on XMPP protocol, a multi-terminal push network of emergency information is constructed to push monitoring information and emergency decision to each push terminal. The experimental results show that the information recognition rate of the proposed method is higher than 99%, and the emergency information push result has high precision.

Partial Disaster Classification Based on Attention Mechanism and Balanced Softmax Loss Function

Partial Disaster Classification Based on Attention Mechanism and Balanced Softmax Loss Function

DeltaVLAD: An efficient optimization algorithm to discriminate speaker embedding for text-independent speaker verification

<abstract> <p>Text-independent speaker verification aims to determine whether two given utterances in open-set task originate from the same speaker or not. In this paper, some ways are explored to enhance the discrimination of embeddings in speaker verification. Firstly, difference is used in the coding layer to process speaker features to form the DeltaVLAD layer. The frame-level speaker representation is extracted by the deep neural network with differential operations to calculate the dynamic changes between frames, which is more conducive to capturing insignificant changes in the voiceprint. Meanwhile, NeXtVLAD is adopted to split the frame-level features into multiple word spaces before aggregating, and subsequently perform VLAD operations in each subspace, which can significantly reduce the number of parameters and improve performance. Secondly, the margin-based softmax loss function and the few-shot learning-based loss function are proposed to be combined for more discriminative speaker embeddings. Finally, for a fair comparison, the experimental results are performed on Voxceleb-1 showing superior performance of speaker verification system and can obtain new state-of-the-art results.</p> </abstract>

Dynamic Training Data Dropout for Robust Deep Face Recognition

Learning with noise is a practically challenging problem in deep face recognition. Despite the success of large margin softmax loss functions, these methods are designed for clean face databases. Considering the inevitable noise in the large scale databases, we first analyze the performance of noise in the training databases. For noise-robust deep face recognition, we propose a dynamic training data dropout (DTDD) method to dynamically filter the noise in the training database and gradually form a stable refined database for model learning. Specifically, we leverage the information provided by the model predictions of accumulated training epochs, which can distinguish regular samples and noise effectively and accurately. The proposed DTDD method is easy and stable for implementation, and can be combined with existing state-of-the-art loss functions and network architectures. Extensive experiments on CASIA-WebFace, VGGFace2, and MS-Celeb-1 M databases empirically demonstrate that our proposed method can robustly train deep face recognition models in the presence of label noise and low quality images.

Deep Siamese Semantic Segmentation Network for PCB Welding Defect Detection

Deep learning has been widely used in recent years for printed circuit board (PCB) defect detection because of its excellent performance. However, deep-learning-based approaches often suffer from the over-fitting problem due to the lack of sufficient training data in real applications. Meanwhile, these approaches still have some challenges to detect these defects with small sizes and irregular shapes. To address these problems, this article has developed a novel deep Siamese semantic segmentation network which integrates the similarity measurement of the Siamese network with the encoder–decoder semantic segmentation network for PCB welding defect detection. This network includes two encoders sharing weighted values, a decoder, and some correlation modules, in which the decoder integrates deep features from two decoders with their feature difference computed by some correlation modules via skipping connections to recover spatial information on multiple output layers, and thus this proposed network can perform PCB welding small defect semantic segmentation. Moreover, via these correlation modules, this proposed network can pay more attention to semantic difference and further alleviate the over-fitting issue because of insufficient defect samples. Finally, we propose a combined loss function which combines the weighted cross-entropy loss, the Lovasz softmax loss, and the weighted precision–recall loss for network training to further improve small defect segmentation and recall improvement. Experimental results demonstrate that the proposed network can be trained on limited training images and achieve high efficiency and outstanding effects for PCB welding small defect segmentation.

Defect segmentation: Mapping tunnel lining internal defects with ground penetrating radar data using a convolutional neural network

This work offers a defect segmentation approach for the nondestructive testing of tunnel lining internal defects using Ground Penetrating Radar (GPR) data. Given GPR synthetic data, it maps the internal defect structure, using a CNN named Segnet coupled with the Lovász softmax loss function, which enhances the accuracy, automation, and efficiency of defect identification. Experiments with both synthetic and actual data show that our innovative method overcomes problems in standard GPR data interpretation. A physical test model with a known defect was developed and manufactured, and GPR data was acquired and analyzed to verify the approach.

Weighted ensemble with angular feature learning for facial expression recognition

Ensemble learning using a set of deep convolutional neural networks (DCNNs) as weak classifiers has become a powerful tool for face expression. Nevertheless, training a DCNNS-based ensemble is not only time consuming but also gives rise to high redundancy due to the nature of DCNNs. In this paper, a novel DCNNs-based ensemble method, named weighted ensemble with angular feature learning (WDEA), is proposed to improve the computational efficiency and diversity of the ensemble. Specifically, the proposed ensemble consists of four parts including input layer, trunk layers, diversity layers and loss fusion. Among them, the trunk layers which are used to extract the local features of face images are shared by diversity layers such that the lower-level redundancy can be largely reduced. The independent branches enable the diversity of the ensemble. Rather than the traditional softmax loss, the angular softmax loss is employed to extract more discriminant deep feature representation. Moreover, a novel weighting technique is proposed to enhance the diversity of the ensemble. Extensive experiments were performed on CK+ and AffectNet. Experimental results demonstrate that the proposed WDEA outperforms existing ensemble learning methods on the recogntion rate and computational efficiency.

A Quadruplet Deep Metric Learning model for imbalanced time-series fault diagnosis

Deep learning based methods are attractive and meaningful in the field of fault diagnosis recently. However, the sample size of different faults may be imbalanced in practical application scenarios, which results in performance degeneration. The optimization of data representation could be an effective way to alleviate such phenomenon. In this paper, a deep learning framework considering time series and distance metric called LSTM-Quadruplet Deep Metric Learning(LSTM-QDM) model is proposed. It maps original space to a feature space where the distribution of imbalance faults is more distinguishable. A novel quadruplet data pair is designed which adds a minor sample from imbalanced classes into traditional data pair. Based on such data pair, a quadruplet loss function is proposed to increase the distance between the imbalanced classes and other classes, and its combination with softmax loss would improve the representation ability and classification performance simultaneously. The proposed data pair and loss function encourage the full mining of imbalanced data in the training process. The experiments are carried out on two open-source datasets including TE and CWRU, and the fault diagnosis performance is validated under different imbalanced conditions. The experimental results indicate that our proposed model is effective and robust in the imbalanced fault diagnosis task.

Two-stream network with phase map for few-shot classification

As an important research direction of deep learning, few-shot classification mainly solves problems of classification when training samples are scarce. Under this circumstance, how to effectively use the information of existing samples has become the key to solve few-shot classification problems. We propose a method to fully mine existing samples information, which can be implemented in the training and testing stages. Specifically, we introduce a two-stream neural network as an embedding model to extract features of the original image and its phase map, and perform fine-grained adaptive feature fusion to them. The algorithm can learn and extract more fully-informed fusion features under the network. Besides, we also introduce Circle Loss in loss function, which has a non-linear decision boundary. Under the joint supervision of Soft-max Loss and Circle Loss, the distinction between different classes of features in embedding space increases. We conduct experiments on miniImageNet, tieredImageNet, and CIFAR-FS, respectively. The accuracy rate is increased by 0.99%, 0.96%, and 0.91%, which proves the effectiveness of our algorithm.

Loss function search for person re-identification

In recent years, person re-identification, which learns discriminative features for the specific person retrieval problem across non-overlapping cameras, has attracted extensive attention. One of the main challenges in person re-identification with deep neural networks is the design of the loss function, which plays a vital role in improving the discrimination of the learned features. However, most existing models utilize the hand-designed loss functions, which are usually sub-optimal and time-consuming. The search spaces of the two existing AutoML-based methods are either too complicated or too simple to include various forms of loss functions. In order to solve the irrationality of the above search spaces, in this paper, we propose a method of AutoML for loss function search named LFS-ReID for person ReID in the framework of the margin-based softmax loss function. Specifically, we first analyze the margin-based softmax loss function and conclude four key properties. Then we carefully design a sampling distribution based on the non-independent truncated Gaussian distributions to sample the loss function, which conforms to the above four properties. Finally, a method based on reinforcement learning is adopted to optimize the sampling distribution dynamically. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on four commonly used datasets.

Pixel Distribution Learning for Vessel Segmentation under Multiple Scales.

In this work we try to address if there is a better way to classify two distributions, rather than using histograms; and answer if we can make a deep learning network learn and classify distributions automatically. These improvements can have wide ranging applications in computer vision and medical image processing. More specifically, we propose a new vessel segmentation method based on pixel distribution learning under multiple scales. In particular, a spatial distribution descriptor named Random Permutation of Spatial Pixels (RPoSP) is derived from vessel images and used as the input to a convolutional neural network for distribution learning. Based on our preliminary experiments we currently believe that a wide network, rather than a deep one, is better for distribution learning. There is only one convolutional layer, one rectified linear layer and one fully connected layer followed by a softmax loss in our network. Furthermore, in order to improve the accuracy of the proposed approach, the RPoSP features are captured at multiple scales and combined together to form the input of the network. Evaluations using standard benchmark datasets demonstrate that the proposed approach achieves promising results compared to the state-of-the-art.

Error Refactor loss based on error analysis in image classification

The loss function is a criterion to evaluate the learning quality of a deep convolutional neural network, which represents the gap between prediction and ground truth. However, as the most commonly used loss function in image classification tasks, Cross-Entropy loss does not encourage the model to distinguish the similarity between features. In this work, the authors investigate inter-class separability of similar features learnt by convolutional networks and propose a loss function called Error Refactor Loss (ER-Loss). ER-Loss is based on the error caused by convolutional networks; it can improve the inter-class separability and is simple to implement and can easily replace the Cross-Entropy loss. Compared with softmax loss, ER-Loss adds a dynamic penalty item which can help ER-Loss monitor the actual situation of model training and adjust the value of the penalty item according to model training. The ER-Loss on CIFAR100 and part of ImageNet ILSVRC 2012 is evaluated and the experimental result showed that the ER-Loss can improve the accuracy of the model.

A-MobileNet: An approach of facial expression recognition

Facial expression recognition (FER) is to separate the specific expression state from the given static image or video to determine the psychological emotions of the recognized object, the realization of the computer's understanding and recognition of facial expressions have fundamentally changed the relationship between human and computer,toachieve better human computer interaction (HCI). In recentyears, FER has attracted widespread attention in the fields of HCI, security, communications and driving, and has become one of the research hotspots. In the mobile Internet era, the need for lightweight networking and real-time performance is growing. In this paper, a lightweight A-MobileNet model is proposed. First, the attention module is introduced into the MobileNetV1 model to enhance the local feature extraction of facial expressions. Then, the center loss and softmax loss are combined to optimize the model parameters to reduce intra-class distance and increase inter-class distance. Compared with the original MobileNet series models, ourmethod significantly improvesrecognition accuracy without increasing the number of model parameters. Compared with others, A- MobileNet model achieves better results on the FERPlus and RAF-DB datasets.

SAR image classification based on multi‐feature fusion decision convolutional neural network

With the success of convolutional neural network (CNN) in the field of image processing and image recognition, an improved fully convolutional neural network (FCNN) based on CNN is proposed. Although FCNN has better performance than CNN, the number of parameters of FCNN has also increased significantly compared to CNN. Consequently, this paper proposes a multi-feature fusion decision convolutional neural network (MFFD-CNN), which use two kinds down-sample methods in the pooling layer, and chooses the concatenate or add operation as the way that consolidates the feature of the convolutional layer with a stride of 2 and the max-pooling layer to execute fusion operations. Meanwhile replaces the traditional softmax loss with an additive margin Softmax (AM-Softmax) loss. The structure, which is based on the multi-target classification experiment results of the MSTAR data set, shows can not only obtain an average correct recognition rate 1.2% higher than FCNN, but also better stability without augmenting the training samples of MSTAR data.

Environmental sound classification using convolution neural networks with different integrated loss functions

AbstractThe hike in the demand for smart cities has gathered the interest of researchers to work on environmental sound classification. Most researchers' goal is to reach the Bayesian optimal error in the field of audio classification. Nonetheless, it is very baffling to interpret meaning from a three‐dimensional audio and this is where different types of spectrograms become effective. Using benchmark spectral features such as mel frequency cepstral coefficients (MFCCs), chromagram, log‐mel spectrogram (LM), and so on audio can be converted into meaningful 2D spectrograms. In this paper, we propose a convolutional neural network (CNN) model, which is fabricated with additive angular margin loss (AAML), large margin cosine loss (LMCL) and a‐softmax loss. These loss functions proposed for face recognition, hold their value in the other fields of study if they are implemented in a systematic manner. The mentioned loss functions are more dominant than conventional softmax loss when it comes to classification task because of its capability to increase intra‐class compactness and inter‐class discrepancy. Thus, with MCAAM‐Net, MCAS‐Net and MCLCM‐Net models, a classification accuracy of 99.60%, 99.43% and 99.37% is achieved on UrbanSound8K dataset respectively without any augmentation. This paper also demonstrates the benefit of stacking features together and the above‐mentioned validation accuracies are achieved after stacking MFCCs and chromagram on the x‐axis. We also visualized the clusters formed by the embedded vectors of test data for further acknowledgement of our results, after passing it through different proposed models. Finally, we show that the MCAAM‐Net model achieved an accuracy of 99.60% on UrbanSound8K dataset, which outperforms the benchmark models like TSCNN‐DS, ADCNN‐5, ESResNet‐Attention, and so on that are introduced over the recent years.

Classification of flower image based on attention mechanism and multi-loss attention network

The accurate classification of flower images is the prerequisite for flower plant management to artificial intelligence, how to use the machine to classify flowers automatically is the current hot issue to be solved. This paper first introduced the principle of attention mechanism and realization of spatial attention mechanism and channel attention mechanism, and then designed the embedding of the spatial attention module and channel attention model in Xception structure based on Xception. Final, the network was optimized by jointing Triplet Loss and Softmax Loss in the network loss layer, to obtain a feature embedding space with high intra-class compactness and inter-class separation. This paper was experimented on two flower image data sets (Oxford 17 flowers and Oxford 102 flowers), the results show that the MLSAN, MLCAN, MLCSAN model proposed in this paper were 0.39%, 0.50%, and 0.72% higher on the Oxford 17 flowers data set​ and 0.52%, 0.63% and 0.85% higher on the data set Oxford 102 flowers data set.

Pedestrian Behavior Recognition Based on Improved Dual-stream Network with Differential Feature in Surveillance Video

In order to improve the pedestrian behavior recognition accuracy of video sequences in complex background, an improved spatial-temporal two-stream network is proposed in this paper. Firstly, the deep differential network is used to replace the temporal-stream network so as to improve the representation ability and extraction efficiency of spatiotemporal features. Then, the improved Softmax loss function based on decision-making level feature fusion mechanism is used to train the model, which can retain the spatiotemporal characteristics of images between different network frames to a greater extent and reflect the action category of pedestrians more realistically. Simulation results show that the proposed improved network achieves 87% recognition accuracy on the self-built infrared dataset, and the computational efficiency is improved by 15.1%.

Intrusion Detection for Industrial Control Systems Based on Open Set Artificial Neural Network

The security of industrial control systems (ICSs) has received a lot of attention in recent years. ICSs were once closed networks. But with the development of IT technologies, ICSs have become connected to the Internet, increasing the potential of cyberattacks. Because ICSs are so tightly linked to human lives, any harm to them could have disastrous implications. As a technique of providing protection, many intrusion detection system (IDS) studies have been conducted. However, because of the complicated network environment and rising means of attack, it is difficult to cover all attack classes, most of the existing classification techniques are hard to deploy in a real environment since they cannot deal with the open set problem. We propose a novel artificial neural network based-methodology to solve this problem. Our suggested method can classify known classes while also detecting unknown classes. We conduct research from two points of view. On the one hand, we use the openmax layer instead of the traditional softmax layer. Openmax overcomes the limitations of softmax, allowing neural networks to detect unknown attack classes. During training, on the other hand, a new loss function termed center loss is implemented to improve detection ability. The neural network model learns better feature representations with the combined supervision of center loss and softmax loss. We evaluate the neural network on NF-BoT-IoT-v2 and Gas Pipeline datasets. The experiments show our proposed method is comparable with the state-of-the-art algorithm in terms of detecting unknown classes. But our method has a better overall classification performance.