Global Discriminative Features Research Articles

Attention-based ConvNeXt with a parallel multiscale dilated convolution residual module for fault diagnosis of rotating machinery

Convolutional Neural Networks have promoted development of the fault diagnosis in the machine prognostics and health management. However, the existing methods have limited applicability under strong noisy conditions. We propose an attention-based ConvNeXt with a parallel multiscale dilated convolution residual module for the rotor fault diagnosis. The parallel multiscale dilated convolution residual module is firstly introduced to filter noise and extract multiscale discriminative features intelligently. The multi-head attention module is then embedded in ConvNeXt for global discriminative features adaptively. Focal Loss is finally implemented for identifying hard-to-classify samples to further improve the diagnostic accuracy. The application results demonstrate the superiority and robustness of our method in two case studies.

Unsupervised Local Discrimination for Medical Images.

Contrastive learning, which aims to capture general representation from unlabeled images to initialize the medical analysis models, has been proven effective in alleviating the high demand for expensive annotations. Current methods mainly focus on instance-wise comparisons to learn the global discriminative features, however, pretermitting the local details to distinguish tiny anatomical structures, lesions, and tissues. To address this challenge, in this paper, we propose a general unsupervised representation learning framework, named local discrimination (LD), to learn local discriminative features for medical images by closely embedding semantically similar pixels and identifying regions of similar structures across different images. Specifically, this model is equipped with an embedding module for pixel-wise embedding and a clustering module for generating segmentation. And these two modules are unified by optimizing our novel region discrimination loss function in a mutually beneficial mechanism, which enables our model to reflect structure information as well as measure pixel-wise and region-wise similarity. Furthermore, based on LD, we propose a center-sensitive one-shot landmark localization algorithm and a shape-guided cross-modality segmentation model to foster the generalizability of our model. When transferred to downstream tasks, the learned representation by our method shows a better generalization, outperforming representation from 18 state-of-the-art (SOTA) methods and winning 9 out of all 12 downstream tasks. Especially for the challenging lesion segmentation tasks, the proposed method achieves significantly better performance. The source codes are publicly available at https://github.com/HuaiChen-1994/LDLearning.

Context-aware and part alignment for visible-infrared person re-identification

Visible-infrared person re-identification (VI-ReID) is a challenging problem of matching a person from visible and infrared modalities. Existing researches adopt the last convolutional layer features of the off-the-shelf backbone network as the representation to refine, which is unable to represent the heterogeneous cross-modality features with discriminant information. In this paper, we propose a novel graph-based aggregation learning network (GALNet) with visual Transformer embedding to mine both multi-layer features and part-level contextual cues for VI-ReID. We propose a novel feature memory module (FMM) to reserve global discriminative features of the low layers with the correlation modeling ability of the graph convolution network (GCN) which is supplemented to the final person representation. To learn more discriminant part-level features, an attentive part aggregation module (PAM) is designed to mine part relationships, leveraging the self-attention mechanism of the Transformer. By fusing these components, the global-level, and part-level discriminant information can be utilized. Extensive experiments on SYSU-MM01 and RegDB benchmarks demonstrate the effectiveness of our proposed methods compared with several state-of-the-art methods.

CariesFG: A fine-grained RGB image classification framework with attention mechanism for dental caries

Dental caries is one of the most prevalent oral diseases, and deep learning methods have been used for caries diagnosis in large populations by leveraging RGB images. The existing attention-based fine-grained image classification methods have the problem of underutilization of features and easy interference by background and irrelevant information. To tackle these issues, we propose a fine-grained RGB image classification framework with attention mechanism for dental caries (CariesFG). Specifically, it consists of 4 components: (1) Multi-Spectral channel Attention Module (MSAM), which can retain the useful frequency components in the feature map. (2) Position Attention Module (PAM), which captures feature dependencies in the spatial dimension. (3) Discriminative Point Selection strategy (DPS), which can find the most discriminative feature points. (4) Graph Convolution and Aggregation module (GCA), which aims to aggregate discriminative feature points at different scales of feature maps. To enhance the ability to extract the discriminative features, PAM and MSAM are integrated into the backbone network to consist of feature extraction networks incorporating attention mechanism. Discriminative feature points at different scales of feature maps are extracted by DPS and aggregated as global discriminative features by GCA. By testing on a caries fine-grained classification dataset, CariesFG achieved an accuracy of 68.36%, an f1-score of 66.77% and a specificity of 84.17%, respectively, significantly outperforming state-of-the-art methods. Moreover, visualization results on attention parts show that CariesFG can effectively learn discriminative features and discriminative parts.

Multi-Panda Tracking

Multi-Panda Tracking (MPT) is a video-based tracking task for panda individuals, which is conducive to the observation and measurement of distribution and status of pandas. Different from tracking general objects such as pedestrians and vehicles, MPT is extremely challenging due to the indistinguishable appearances and diversified postures of pandas. In this case, existing tracking methods cannot appropriately tackle with the excessive occlusion between different panda individuals, hence suffering from identity switch, missing and inaccurate detections. To address these problems, we propose a simple yet effective MPT framework in the tracking-by-detection paradigm, which is benefited both from a short-term prediction filtering module and a discriminative feature learning network. In particular, the short-term prediction filtering module introduces similarity learning to enhance the temporal consistency among detections, which is capable of supplementing the missing detections and discarding false positive detections. Besides, the discriminative feature learning network leverages a two-branch network to learn both local and global discriminative features, so as to distinguish different panda individuals with a very similar appearance with a subtle difference. To evaluate the proposed method, we annotate a large-scale MPT dataset, named PANDA2021, which is particularly challenging due to the similar appearance and dramatic occlusion between panda individuals. Experiments on PANDA2021 demonstrate that the proposed MPT method significantly outperforms the competing methods. Moreover, experimental results on pedestrian tracking dataset MOT16 further demonstrate that the proposed MPT method achieves comparative performance with competing methods.

Transmittance Surface Detection and Material Identification Using Multitask ViT-SIFT Fusion

The automatic detection of transmittance surfaces (transparency and translucency) and material class identification (glass, plastic, and so on) from an image is crucial for applications, including domestic service robotics and handling fragile instruments in experimental or industrial settings. However, this is a difficult task due to the lack of expressive textures in such materials. Although previous approaches for detecting transmittance surfaces propose handcrafted features, they disregard global material information and are limited to a single item (usually glass). Material or object recognition algorithms based on convolutional neural networks (CNNs) focus on optimizing performance that requires large labeled datasets. These algorithms are better for detecting opaque objects because large architectures tend to generalize the texture from an image’s background layer. Furthermore, creating large-scale off-the-shelf dataset labeled with transmittance surface is an extremely tedious task. Thus, we propose a simple yet effective and efficient learning model that combines a shallow multitask vision transformer (ViT) with a scale-invariant feature transform (SIFT) and a shallow CNN as the backbone network. SIFT and ViT are used to capture local and global discriminative features, respectively, and their fusion can increase the performance on a small dataset. The backbone network is utilized to introduce the desirable properties of CNN. Using limited training data, this model detects transmittance surfaces and determines their material type. We conducted tests on a new benchmark dataset containing a variety of transmissive materials and three publicly available datasets for robust comparison. Results show that the proposed model achieved a new state-of-the-art accuracy of 81.78% and 77.21% on our new dataset for transmittance surface recognition and material class identification, respectively.

SPD matrices representing artery anatomy for first-pass effect prediction by aggregated networks with multi-scale attentions

ABSTRACT First-pass Effect(FPE) for Endovascular Therapy(EVT) is associated with good clinical outcome (mRS2) of patients with Acute Ischaemic Stroke (AIS). A rapid and accurate prediction of FPE before EVT can help neurointerventionists plan the procedure and avoid delays in restoration of cerebral blood flow. However, there are rare studies focused on FPE prediction on arterial vessel anatomy immediately. The intractable difficulty lies in extracting discriminative features to represent a wide variety of vessels with irregular vessel shapes. In this paper, we propose a Symmetric Positive Definite(SPD) matrix-based feature representation extracted from the centerline of arterial vessels, encoding the global discriminative information over the artery for predicting FPE. Subsequently, a collaborative network of multi-scale Convolutional Neural Network(CNN) and Multiple Layer Perception (MLP) with specific attentions is developed. Specifically, the CNN is used for capturing the features among the multi-scale local neighbours of the curve. MLPs are used for capturing more prominent global discriminative features at different scales. The attention mechanism is used to better filter and extract the useful information for feature fusion. Quantitative experimental results demonstrate that our proposed method is able to predict FPE accurately, outperforming the manually defined features and traditional machine learning-based methods in this task, regarding the metrics of AUC, precision, sensitivity, specificity and accuracy.

Discriminative Feature Mining and Enhancement Network for Low-Resolution Fine-Grained Image Recognition

Existing fine-grained image recognition methods are difficult to learn complete discriminative features from low-resolution (LR) data, because the original subtle inter-class distinctions become slimmer with the reduction of the image resolution. Besides, existing methods of LR fine-grained image recognition and general LR image recognition only consider the restoration and extraction of global discriminative features, ignoring unreliable local fine-grained details can be detrimental to final recognition. To address the above problems, we propose a multi-tasking framework, discriminative feature mining and enhancement network (DME-Net), for the LR fine-grained image recognition task, which aims to capture the reliable object descriptions from macro and micro perspectives, respectively. Macroscopically, we train the framework’s ability to recover and extract global discriminative features based on the whole images. Microscopically, we purposefully reinforce the framework’s ability to repair and capture the local discriminative details on the mined informative parts. To precisely excavate the most potential parts, we design an informative part mining (IPM) module, in which we firstly employ a part generation layer to predict several part masks that focus on different discriminative parts under the guidance of discrepancy loss and discriminant loss. Then we introduce a part selection (PS) submodule to further screen out a group of most informative parts from the predicted part masks according to their corresponding scores, which measure the semantic correlation degree of each part to the others. Experimental results on three benchmark datasets and one retail product dataset consistently show that our proposed framework can significantly boost the performance of the baseline model. Besides, extensive ablation studies are conducted, which further prove the effectiveness of each component of our designs.

Global contextual residual convolutional neural networks for motor fault diagnosis under variable-speed conditions

Convolutional neural networks, with a powerful ability for feature representation, have made vast inroads into motor fault diagnosis. However, most of the existing CNN models cannot favorably handle the data generated in variable-speed scenarios. First, the continuous irregular fluctuation of the motor makes the time domain interval between two adjacent fault pulses change continuously. Secondly, due to the complex transmission path of the signal under unstable conditions, the noise distribution is complex. To address this problem, a global contextual residual convolutional neural network is proposed. The major novelties fall into three aspects. First, to make full use of the features from all intermediate layers and explore multiscale information, a new hierarchical structure is adopted in the CNN model. Second, since different features are of different importance for fault detection tasks, the global context module is explored to guide the model to pay more attention to global discriminant features. Third, the features learned by the network can either promote each other or contradict each other, so a multi-feature fusion layer is introduced to integrate these features adaptively. Case studies using the benchmark motor dataset and the industrial motor bearing dataset are performed to validate the superiority of the GC-ResCNN.

A CNN-Assisted Enhanced Audio Signal Processing for Speech Emotion Recognition.

Speech is the most significant mode of communication among human beings and a potential method for human-computer interaction (HCI) by using a microphone sensor. Quantifiable emotion recognition using these sensors from speech signals is an emerging area of research in HCI, which applies to multiple applications such as human-reboot interaction, virtual reality, behavior assessment, healthcare, and emergency call centers to determine the speaker’s emotional state from an individual’s speech. In this paper, we present major contributions for; (i) increasing the accuracy of speech emotion recognition (SER) compared to state of the art and (ii) reducing the computational complexity of the presented SER model. We propose an artificial intelligence-assisted deep stride convolutional neural network (DSCNN) architecture using the plain nets strategy to learn salient and discriminative features from spectrogram of speech signals that are enhanced in prior steps to perform better. Local hidden patterns are learned in convolutional layers with special strides to down-sample the feature maps rather than pooling layer and global discriminative features are learned in fully connected layers. A SoftMax classifier is used for the classification of emotions in speech. The proposed technique is evaluated on Interactive Emotional Dyadic Motion Capture (IEMOCAP) and Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) datasets to improve accuracy by 7.85% and 4.5%, respectively, with the model size reduced by 34.5 MB. It proves the effectiveness and significance of the proposed SER technique and reveals its applicability in real-world applications.

Learning Structural Representations via Dynamic Object Landmarks Discovery for Sketch Recognition and Retrieval.

State-of-the-art methods on sketch classification and retrieval are based on deep convolutional neural network to learn representations. Although deep neural networks have the ability to model images with hierarchical representations by convolution kernels, they can not automatically extract the structural representations of object categories in a human-perceptible way. Furthermore, sketch images usually have large scale visual variations caused by the styles of drawing or viewpoints, which make it difficult to develop generalized representations using the fixed computational mode of convolutional kernel. In this paper, our aim is to address the problem of fixed computational mode in feature extraction process without extra supervision. We propose a novel architecture to dynamically discover the object landmarks and learn the discriminative structural representations. Our model is composed of two components: a representative landmark discovering module that localizes the key points on the object, and a category-aware representation learning module that develops the category-specific features. Specifically, we develop a structure-aware offset layer to dynamically localize the representative landmarks, which is optimized based on the category labels without extra supervision. After that, a diversity branch is introduced to extract the global discriminative features for each category. Finally, we employ a multi-task loss function to develop an end-to-end trainable architecture. At testing time, we fuse all the predictions with different number of landmarks to achieve the final results. Through extensive experiments, we compare our model with several state-of-the-art methods on two challenging datasets TU-Berlin and Sketchy for sketch classification and retrieval, and the experimental results demonstrate the effectiveness of our proposed model.

An improved hybrid approach to face recognition by fusing local and global discriminant features

This paper presents a novel hybrid method which exploits both local and global discriminative features by a fusion for face representation and recognition. The whole face images are divided into a number of non-overlapping sub-regions to extract local discriminant features. The global discriminant features are extracted from the whole face images. The PCA and FLD methods are applied on the fused feature vector to extract lower dimensional discriminant features. The simulation results of the proposed method on the CMU PIE, FERET and AR face databases have been compared with other approaches and have demonstrated consistently improved performances.

Discriminative and Non-User Specific Binary Biometric Representation via Linearly-Separable SubCode Encoding-based Discretization

Biometric discretization is a process of transforming continuous biometric features of an identity into a binary bit string. This paper mainly focuses on improving the global discretization method ? a discretization method that does not base on information specific to each user in bitstring extraction, which appears to be important in applications that prioritize strong security provision and strong privacy protection. In particular, we demonstrate how the actual performance of a global discretization could further be improved by embedding a global discriminative feature selection method and a Linearly Separable Subcode-based encoding technique. In addition, we examine a number of discriminative feature selection measures that can reliably be used for such discretization. Lastly, encouraging empirical results vindicate the feasibility of our approach.