A Multi-feature Fusion and SSAE-Based Deep Network for Image Semantic Recognition

In view of the scene’s complexity and diversity in scene classification, this paper makes full use of the contextual semantic relationships between the objects to describe the visual attention regions of the scenes and combines with the deep convolution neural networks, so that a scene classification model using visual attention and deep networks is constructed. Firstly, the visual attention regions in the scene image are marked by using the context-based saliency detection algorithm. Then, the original image and the visual attention region detection image are superimposed to obtain a visual attention region enhancement image. Furthermore, the deep convolution features of the original image, the visual attention region detection image, and the visual attention region enhancement image are extracted by using the deep convolution neural networks pretrained on the large-scale scene image dataset Places. Finally, the deep visual attention features are constructed by using the multilayer deep convolution features of the deep convolution networks, and a classification model is constructed. In order to verify the effectiveness of the proposed model, the experiments are carried out on four standard scene datasets LabelMe, UIUC-Sports, Scene-15, and MIT67. The results show that the proposed model improves the performance of the classification well and has good adaptability.

Deep learning algorithms have shown significant performance improvements for facial expression recognition (FER). Most deep learning-based methods, however, focus more attention on spatial appearance features for classification, discarding much useful temporal information. In this work, we present a novel framework that jointly learns spatial features and temporal dynamics for FER. Given the image sequence of an expression, spatial features are extracted from each frame using a deep network, while the temporal dynamics are modeled by a convolutional network, which takes a pair of consecutive frames as input. Finally, the framework accumulates clues from fused features by a BiLSTM network. In addition, the framework is end-to-end learnable, and thus temporal information can be adapted to complement spatial features. Experimental results on three benchmark databases, CK+, Oulu-CASIA and MMI, show that the proposed framework outperforms state-of-the-art methods.

Seeing through disguise: Getting to know you with a deep convolutional neural network

With the development of theories and technologies in medical imaging, most of the tumors can be detected in the early stage. However, the nature of ovarian cysts lacks accurate judgement, leading to that many patients with benign nodules still need Fine Needle Aspiration (FNA) biopsies or surgeries, increasing the physical pain and mental pressure of patients as well as unnecessary medical health care costs. Therefore, we present an image diagnosis system for classifying the ovarian cysts in color ultrasound images, which novelly applies the image features fused by both high-level features from deep learning network and low-level features from texture descriptor. Firstly, the ultrasound images are enhanced to improve the quality of training data set and the rotation invariant uniform local binary pattern (ULBP) features are extracted from each of the images as the low-level texture features. Then the high-level deep features extracted by the fine-tuned GoogLeNet neural network and the low-level ULBP features are normalized and cascaded as one fusion feature that can represent both the semantic context and the texture patterns distributed in the image. Finally, the fusion features are input to the Cost-sensitive Random Forest classifier to classify the images into "malignant" and "benign". The high-level features extracted by the deep neural network from the medical ultrasound image can reflect the visual features of the lesion region, while the low-level texture features can describe the edges, direction and distribution of intensities. Experimental results indicate that the combination of the two types of features can describe the differences between the lesion regions and other regions, and the differences between lesions regions of malignant and benign ovarian cysts.

Breast tumor is a fatal threat to the health of women. Ultrasound (US) is a common and economical method for the diagnosis of breast cancer. Breast imaging reporting and data system (BI-RADS) category 4 has the highest false-positive value of about 30% among five categories. The classification task in BI-RADS category 4 is challenging and has not been fully studied. This work aimed to use convolutional neural networks (CNNs) for breast tumor classification using B-mode images in category 4 to overcome the dependence on operator and artifacts. Additionally, this work intends to take full advantage of morphological and textural features in breast tumor US images to improve classification accuracy. First, original US images coming directly from the hospital were cropped and resized. In 1385 B-mode US BI-RADS category 4 images, the biopsy eliminated 503 samples of benign tumor and left 882 of malignant. Then, K-means clustering algorithm and entropy of sliding windows of US images were conducted. Considering the diversity of different characteristic information of malignant and benign represented by original B-mode images, K-means clustering images and entropy images, they are fused in a three-channel form multi-feature fusion images dataset. The training, validation, and test sets are 969, 277, and 139. With transfer learning, 11 CNN models including DenseNet and ResNet were investigated. Finally, by comparing accuracy, precision, recall, F1-score, and area under curve (AUC) of the results, models which had better performance were selected. The normality of data was assessed by Shapiro-Wilk test. DeLong test and independent t-test were used to evaluate the significant difference of AUC and other values. False discovery rate was utilized to ultimately evaluate the advantages of CNN with highest evaluation metrics. In addition, the study of anti-log compression was conducted but no improvement has shown in CNNs classification results. With multi-feature fusion images, DenseNet121 has highest accuracy of 80.22±1.45% compared to other CNNs, precision of 77.97±2.89% and AUC of 0.82±0.01. Multi-feature fusion improved accuracy of DenseNet121 by 1.87% from classification of original B-mode images (p<0.05). The CNNs with multi-feature fusion show a good potential of reducing the false-positive rate within category 4. The work illustrated that CNNs and fusion images have the potential to reduce false-positive rate in breast tumor within US BI-RADS category 4, and make the diagnosis of category 4 breast tumors to be more accurate and precise.

Hyperspectral remote sensing image classification based on residual generative Adversarial Neural Networks

In this paper, we propose a new image super-resolution (SR) approach based on a convolutional neural network (CNN), which jointly learns the feature extraction, upsampling, and high-resolution (HR) reconstruction modules, yielding a completely end-to-end trainable deep CNN. However, directly training such a deep network in an end-to-end fashion is challenging, which takes a longer time to converge and may lead to sub-optimal results. To address this issue, we propose to jointly train an ensemble of deep and shallow networks. The shallow network with weaker learning capability restores the main structure of the image content, while the deep network with stronger representation power captures the high-frequency details. Since the shallow network is much easier to optimize, it significantly lowers the difficulty of deep network optimization during joint training. To further ensure more accurate restoration of HR images, the high-frequency details are reconstructed in a multi-scale manner to simultaneously incorporate both short- and long-range contextual information. The proposed method is extensively evaluated on widely adopted data sets and compares favorably against state-of-the-art methods. In-depth ablation studies are conducted to verify the contributions of different network designs to image SR, providing additional insights for future research.

Text recognition in natural scene images is a challenging problem in computer vision. Different than the optical character recognition (OCR), text recognition in natural scene images is more complex due to variations in text size, colors, fonts, orientations, complex backgrounds, occlusion, illuminations and uneven lighting conditions. In this paper, we propose a segmentation-free method based on a deep convolutional recurrent neural network to solve the problem of cursive text recognition, particularly focusing on Urdu text in natural scenes. Compared to the non-cursive scripts, Urdu text recognition is more complex due to variations in the writing styles, several shapes of the same character, connected text, ligature overlapping, stretched, diagonal and condensed text. The proposed model gets a whole word image as an input without pre-segmenting into individual characters, and then transforms into the sequence of the relevant features. Our model is based on three components: a deep convolutional neural network (CNN) with shortcut connections to extract and encode the features, a recurrent neural network (RNN) to decode the convolutional features, and a connectionist temporal classification (CTC) to map the predicted sequences into the target labels. To increase the text recognition accuracy further, we explore deeper CNN architectures like VGG-16, VGG-19, ResNet-18 and ResNet-34 to extract more appropriate Urdu text features, and compare the recognition results. To conduct the experiments, a new large-scale benchmark dataset of cropped Urdu word images in natural scenes is developed. The experimental results show that the proposed deep CRNN network with shortcut connections outperform than other network architectures. The dataset is publicly available and can be downloaded from <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://data.mendeley.com/datasets/k5fz57zd9z/1</uri> .

The classification of high-resolution (HR) synthetic aperture radar (SAR) images is of great importance for SAR scene interpretation and application. However, the presence of intricate spatial structural patterns and complex statistical nature makes SAR image classification a challenging task, especially in the case of limited labeled SAR data. This paper proposes a novel HR SAR image classification method, using a multi-scale deep feature fusion network and covariance pooling manifold network (MFFN-CPMN). MFFN-CPMN combines the advantages of local spatial features and global statistical properties and considers the multi-feature information fusion of SAR images in representation learning. First, we propose a Gabor-filtering-based multi-scale feature fusion network (MFFN) to capture the spatial pattern and get the discriminative features of SAR images. The MFFN belongs to a deep convolutional neural network (CNN). To make full use of a large amount of unlabeled data, the weights of each layer of MFFN are optimized by unsupervised denoising dual-sparse encoder. Moreover, the feature fusion strategy in MFFN can effectively exploit the complementary information between different levels and different scales. Second, we utilize a covariance pooling manifold network to extract further the global second-order statistics of SAR images over the fusional feature maps. Finally, the obtained covariance descriptor is more distinct for various land covers. Experimental results on four HR SAR images demonstrate the effectiveness of the proposed method and achieve promising results over other related algorithms.

Deep learning (also known as deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and nonlinear transformations. Deep learning has been characterized as a class of machine learning algorithms with the following characteristics [257]: • They use a cascade of many layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. The algorithms may be supervised or unsupervised and applications include pattern analysis (unsupervised) and classification (supervised). • They are based on the (unsupervised) learning of multiple levels of features or representations of the data. Higher-level features are derived from lower-level features to form a hierarchical representation. • They are part of the broader machine learning field of learning representations of data. • They learn multiple levels of representations that correspond to different levels of abstraction; the levels form a hierarchy of concepts. These definitions have in common: multiple layers of nonlinear processing units and the supervised or unsupervised learning of feature representations in each layer, with the layers forming a hierarchy from low-level to high-level features. Various deep learning architectures such as deep neural networks, convolutional deep neural networks, deep belief networks (DBN), and recurrent neural networks have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition, and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. In this chapter we start in Section 7.1 with an introduction, giving a brief history of this field, the relevant literature, and its applications. Then we study some basic concepts of deep learning such as convolutional neural networks, recurrent neural networks, backpropagation algorithm, restricted Boltzmann machines, and deep learning networks in Section 7.2. Then we illustrate three examples for Apache Spark implementation for mobile big data (MBD), user moving pattern extraction, and combination with nonparametric Bayesian learning, respectively, in Sections 7.3 through 7.5. Finally, we have summary in Section 7.6.

Recently, the fusion of hyperspectral and light detection and ranging (LiDAR) data has obtained a great attention in the remote sensing community. In this paper, we propose a new feature fusion framework using deep neural network (DNN). The proposed framework employs a novel 3D convolutional neural network (CNN) to extract the spectral-spatial features of hyperspectral data, a deep 2D CNN to extract the elevation features of LiDAR data, and then a fully connected deep neural network to fuse the extracted features in the previous CNNs. Through the aforementioned three deep networks, one can extract the discriminant and invariant features of hyperspectral and LiDAR data. At last, logistic regression is used to produce the final classification results. The experimental results reveal that the proposed deep fusion model provides competitive results. Furthermore, the proposed deep fusion idea opens a new window for future research.

Image retrieval is one of the most interesting and fastest growing research areas in all fields. It is an effective and efficient tool for managing large image databases. In most Content-Based Image Retrieval (CBIR) systems, an image is represented by a set of low-level visual features; hence a direct correlation with high-level semantic information will be absent. Therefore, a gap exists between high-level information and low-level features, which is the main reason that hinders the improvement of the image retrieval accuracy. In this work, main focus is on the semantic based image retrieval system using Gray Level Co-occurrence Matrix (GLCM) for texture feature extraction. Based on the texture features, semantic interpretations are given to the extracted textures. The images are retrieved according to user satisfaction and thereby reduce the semantic gap between low level features and high level features.

Fusing features from different feature descriptors or different convolutional layers can improve the understanding of scene and enhance the classification accuracy. In this article, we propose a hierarchical deep texture feature fusion network, abbreviated as HDTFF-Net, aiming to improve the classification accuracy of high-resolution remote sensing scene classification. The proposed HDTFF-Net can effectively combine the shallow texture information from manual features and the deep texture information by convolutional neural networks (CNNs). First, for deeply excavating the multiscale and multidirectional shallow texture features in images, an improved Wavelet feature extraction module and a Gabor feature extraction module are designed by fully fusing the structural features into the backbone neural network. Then, to make the output texture features more discriminative and interpretative, we incorporate the above texture feature extraction modules into traditional CNNs (Tra-CNNs), and design two improved deep networks, namely Wave-CNN and Gabor-CNN. Finally, according to the Dempster-Shafer evidence theory, the designed Wave-CNN and Gabor-CNN are fused with the Tra-CNN by a decision-level fusion strategy, which can effectively capture the deep texture features by different feature descriptors and improve the classification performance. Experiments on high-resolution remote sensing images demonstrate the effectiveness of the proposed HDTFF-Net, and verify that it can greatly improve the classification performance.

Hyperspectral image (HSI) classification accuracy has been greatly improved by employing deep learning. The current research mainly focuses on how to build a deep network to improve the accuracy. However, these networks tend to be more complex and have more parameters, which makes the model difficult to train and easy to overfit. Therefore, we present a lightweight deep convolutional neural network (CNN) model called S2FEF-CNN. In this model, three S2FEF blocks are used for the joint spectral–spatial features extraction. Each S2FEF block uses 1D spectral convolution to extract spectral features and 2D spatial convolution to extract spatial features, respectively, and then fuses spectral and spatial features by multiplication. Instead of using the full connected layer, two pooling layers follow three blocks for dimension reduction, which further reduces the training parameters. We compared our method with some state-of-the-art HSI classification methods based on deep network on three commonly used hyperspectral datasets. The results show that our network can achieve a comparable classification accuracy with significantly reduced parameters compared to the above deep networks, which reflects its potential advantages in HSI classification.

We propose a technique for producing `visual explanations' for decisions from a large class of Convolutional Neural Network (CNN)-based models, making them more transparent. Our approach - Gradient-weighted Class Activation Mapping (Grad-CAM), uses the gradients of any target concept (say logits for `dog' or even a caption), flowing into the final convolutional layer to produce a coarse localization map highlighting the important regions in the image for predicting the concept. Unlike previous approaches, Grad- CAM is applicable to a wide variety of CNN model-families: (1) CNNs with fully-connected layers (e.g. VGG), (2) CNNs used for structured outputs (e.g. captioning), (3) CNNs used in tasks with multi-modal inputs (e.g. visual question answering) or reinforcement learning, without architectural changes or re-training. We combine Grad-CAM with existing fine-grained visualizations to create a high-resolution class-discriminative visualization, Guided Grad-CAM, and apply it to image classification, image captioning, and visual question answering (VQA) models, including ResNet-based architectures. In the context of image classification models, our visualizations (a) lend insights into failure modes of these models (showing that seemingly unreasonable predictions have reasonable explanations), (b) outperform previous methods on the ILSVRC-15 weakly-supervised localization task, (c) are more faithful to the underlying model, and (d) help achieve model generalization by identifying dataset bias. For image captioning and VQA, our visualizations show even non-attention based models can localize inputs. Finally, we design and conduct human studies to measure if Grad-CAM explanations help users establish appropriate trust in predictions from deep networks and show that Grad-CAM helps untrained users successfully discern a `stronger' deep network from a `weaker' one even when both make identical predictions. Our code is available at https: //github.com/ramprs/grad-cam/ along with a demo on CloudCV [2] and video at youtu.be/COjUB9Izk6E.