Shallow Convolutional Neural Network Architecture Research Articles

Breast cancer detection: Shallow convolutional neural network against deep convolutional neural networks based approach.

Introduction: Of all the cancers that afflict women, breast cancer (BC) has the second-highest mortality rate, and it is also believed to be the primary cause of the high death rate. Breast cancer is the most common cancer that affects women globally. There are two types of breast tumors: benign (less harmful and unlikely to become breast cancer) and malignant (which are very dangerous and might result in aberrant cells that could result in cancer). Methods: To find breast abnormalities like masses and micro-calcifications, competent and educated radiologists often examine mammographic images. This study focuses on computer-aided diagnosis to help radiologists make more precise diagnoses of breast cancer. This study aims to compare and examine the performance of the proposed shallow convolutional neural network architecture having different specifications against pre-trained deep convolutional neural network architectures trained on mammography images. Mammogram images are pre-processed in this study's initial attempt to carry out the automatic identification of BC. Thereafter, three different types of shallow convolutional neural networks with representational differences are then fed with the resulting data. In the second method, transfer learning via fine-tuning is used to feed the same collection of images into pre-trained convolutional neural networks VGG19, ResNet50, MobileNet-v2, Inception-v3, Xception, and Inception-ResNet-v2. Results: In our experiment with two datasets, the accuracy for the CBIS-DDSM and INbreast datasets are 80.4%, 89.2%, and 87.8%, 95.1% respectively. Discussion: It can be concluded from the experimental findings that the deep network-based approach with precise tuning outperforms all other state-of-the-art techniques in experiments on both datasets.

Surface crack detection using deep learning with shallow CNN architecture for enhanced computation

Surface cracks on the concrete structures are a key indicator of structural safety and degradation. To ensure the structural health and reliability of the buildings, frequent structure inspection and monitoring for surface cracks is important. Surface inspection conducted by humans is time-consuming and may produce inconsistent results due to the inspectors’ varied empirical knowledge. In the field of structural health monitoring, visual inspection of surface cracks on civil structures using deep learning algorithms has gained considerable attention. However, these vision-based techniques require high-quality images as inputs and depend on high computational power for image classification. Thus, in this study, shallow convolutional neural network (CNN)-based architecture for surface concrete crack detection is proposed. LeNet-5, a well-known CNN architecture, is optimized and trained for image classification using 40,000 images in the Middle East Technical University (METU) dataset. To achieve maximum accuracy for crack detection with minimum computation, the hyperparameters of the proposed model were optimized. The proposed model enables the employment of deep learning algorithms using low-power computational devices for a hassle-free monitoring of civil structures. The performance of the proposed model is compared with those of various pretrained deep learning models, such as VGG16, Inception, and ResNet. The proposed shallow CNN architecture was found to achieve a maximum accuracy of 99.8% in the minimum computation. Better hyperparameter optimization in CNN architecture results in higher accuracy even with a shallow layer stack for enhanced computation. The evaluation results confirm the incorporation of the proposed method with autonomous devices, such as unmanned aerial vehicle, for real-time inspection of surface crack with minimum computation.

On Convolutional Neural Networks for Chest X-ray Classification

Pneumonia is an infectious disease accounting for one fifth of the deaths of children under five worldwide. It is also a cause for adult hospital admissions with mortality rate of almost 25% in patients over 75 years. Pneumonia is curable and mortality can be prevented if it is diagnosed. Preferred diagnostic technique is chest X-ray image examination. The lack of radiology equipment of trained clinicians reduces the chance of the majority of people to be properly diagnosed. In the paper we propose a shallow convolutional neural network architecture, further fine-tuned with chose of the Adam optimizer which is split tested in the current experimental work. We showed that the automatic detection of pneumonia in chest X-ray images is possible with accuracy higher than 90% using 3 blocks of 2 convolutional 2D layers with max-pooling for feature extraction and a flatten output block comprising two dense layers, trained over 10 epochs. Therefore, the application for CNN is a viable solution as a supplement to the decision-making process of pneumonia X-ray diagnostics.

Detection of Computer Graphics Using Attention-Based Dual-Branch Convolutional Neural Network from Fused Color Components.

With the development of 3D rendering techniques, people can create photorealistic computer graphics (CG) easily with the advanced software, which is of great benefit to the video game and film industries. On the other hand, the abuse of CGs has threatened the integrity and authenticity of digital images. In the last decade, several detection methods of CGs have been proposed successfully. However, existing methods cannot provide reliable detection results for CGs with the small patch size and post-processing operations. To overcome the above-mentioned limitation, we proposed an attention-based dual-branch convolutional neural network (AD-CNN) to extract robust representations from fused color components. In pre-processing, raw RGB components and their blurred version with Gaussian low-pass filter are stacked together in channel-wise as the input for the AD-CNN, which aims to help the network learn more generalized patterns. The proposed AD-CNN starts with a dual-branch structure where two branches work in parallel and have the identical shallow CNN architecture, except that the first convolutional layer in each branch has various kernel sizes to exploit low-level forensics traces in multi-scale. The output features from each branch are jointly optimized by the attention-based fusion module which can assign the asymmetric weights to different branches automatically. Finally, the fused feature is fed into the following fully-connected layers to obtain final detection results. Comparative and self-analysis experiments have demonstrated the better detection capability and robustness of the proposed detection compared with other state-of-the-art methods under various experimental settings, especially for image patch with the small size and post-processing operations.

A divide-and-conquer strategy for facial landmark detection using dual-task CNN architecture

In this paper, we propose a novel deep learning-based framework for facial landmark detection. This framework takes as input face image returned by a face detector (Faster R-CNN) and generates as output a set of landmarks positions. Prior CNN-based methods often select randomly small local patches to predict an initial guess of landmarks locations. One issue with these local patches is that the adjacent landmarks might share the same regions due to the overlapping, thus, they might not convey precise information of each individual landmark. By contrast, our approach formulates this problem as a divide-conquer search for facial patches using CNN architecture in a hierarchy, where the input face image is recursively split into two cohesive non-overlapped subparts until each one contains only the region around the expected landmark. To attain better division of face topology, the search is carried out in a structured coarse-to-fine manner, where a learned hierarchical model of the face defining the granularity of each division level is introduced. We also propose a cascaded regressor to detect and refine the position of the individual landmark in each predicted non-overlapped patch. We adopt a carefully designed shallow CNN architecture so that to improve real-time performance. In addition, unlike previous cascaded methods, our regressor does not require auxiliary input such as initial landmarks locations. Extensive experiments on several challenging datasets (including MTFL, AFW, AFLW, COFW, 300W, and 300VW) show that our approach is particularly impressive in the unconstrained scenarios where it outperforms prior arts in both accuracy and efficiency.

A deep learning CNN architecture applied in smart near-infrared analysis of water pollution for agricultural irrigation resources

Water is a natural resource for agricultural irrigation. Recycling use of water is important in terms of resource conservation and is good for sustainable development of the ecological environment. The wastewater from daily living and industrial production contains various chemicals that are supposed as pollutants leading to the decline of water quality. For the demand of water protection and recycling, the assessment of water pollution level should be evaluated. An effective scientific technique is required for rapid detection of water pollution. Near-infrared (NIR) spectroscopy is a modern technology suitable for rapid detection of agricultural targets. For monitoring the agricultural water resource, the NIR modeling methods are required to be smart and artificially controlled to solve the issues when we confront a considerable number of data or a dynamic situation. In this study, an improved convolutional neural network (CNN) architecture was designed for a deep calibration on the NIR data. The architecture is shallow, simply constructed with one convolution layer and one pooling layer. The decision tree algorithm was employed in the pooling layer for extracting the informative features in a data driven manner. The CNN architecture was trained by combined tuning of multiple parameters in different layers. The convolution filters, the decision tree branches and the hidden neurons in the fully connected layer were automatically adaptive with fidelity to the measured data. A CNN calibration model for NIR quantitatively determination of water pollution level was then established and optimized in deep learning mode, and eventually improved the NIR prediction accuracy. Prospectively, the designed shallow CNN architecture is feasible to be used for establishing intelligent spectroscopic models for evaluating the level of water pollution, and is expected to provide smart technical support in dealing with the issues of water recycling and conservation for agricultural cultivation.

Leveraging Shape, Reflectance and Albedo From Shading for Face Presentation Attack Detection

Presentation attack detection is a challenging problem that aims at exposing an impostor user seeking to deceive the authentication system. In facial biometrics systems, this kind of attack is performed using a photograph, video, or 3D mask containing the biometric information of a genuine identity. In this paper, we propose a novel approach to detecting face presentation attacks based on intrinsic properties of the scene such as albedo, depth, and reflectance properties of the facial surfaces, which were recovered through a shape-from-shading (SfS) algorithm. To extract meaningful patterns from the different maps obtained with the SfS algorithm, we designed a novel shallow CNN architecture for learning features useful to the presentation attack detection (PAD). We performed several experiments considering the intra- and inter-dataset evaluation protocols. The obtained results showed the effectiveness of the proposed method considering several types of photo- and video-based presentation attacks, and in the cross-sensor scenario, besides achieving competitive results for the inter-dataset evaluation protocol.

A Convolutional Neural Network Based Auto Features Extraction Method for Tea Classification with Electronic Tongue

Feature extraction is a key part of the electronic tongue system. Almost all of the existing features extraction methods are “hand-crafted”, which are difficult in features selection and poor in stability. The lack of automatic, efficient and accurate features extraction methods has limited the application and development of electronic tongue systems. In this work, a convolutional neural network-based auto features extraction strategy (CNN-AFE) in an electronic tongue (e-tongue) system for tea classification was proposed. First, the sensor response of the e-tongue was converted to time-frequency maps by short-time Fourier transform (STFT). Second, features were extracted by convolutional neural network (CNN) with time-frequency maps as input. Finally, the features extraction and classification results were carried out under a general shallow CNN architecture. To evaluate the performance of the proposed strategy, experiments were held on a tea database containing 5100 samples for five kinds of tea. Compared with other features extraction methods including features of raw response, peak-inflection point, discrete cosine transform (DCT), discrete wavelet transform (DWT) and singular value decomposition (SVD), the proposed model showed superior performance. Nearly 99.9% classification accuracy was obtained and the proposed method is an approximate end-to-end features extraction and pattern recognition model, which reduces manual operation and improves efficiency.

Recognizing Facial Expressions Using a Shallow Convolutional Neural Network

Generally, facial expressions could be classified into two categories: static facial expressions and micro-expressions. There are many promising applications of facial expression recognition, such as pain detection, lie detection, and babysitting. Traditional convolutional neural network (CNN)-based methods suffer from two critical problems when they are adopted to recognize micro-expressions. First, they are usually dependent on very deep architectures that overfit on small datasets. However, reliable expressions are relatively difficult to collect and relevant datasets are usually relatively small. Second, for micro-expressions, these methods usually neglect the temporal redundancy of micro-expressions which could be utilized to reduce the temporal complexity. In this paper, we propose a shallow CNN (SHCNN) architecture with only three layers to classify static expressions and micro-expressions simultaneously without big training datasets. To better explain the functionality of our SHCNN architecture, we improve the saliency maps by introducing a shrinkage factor after studying the vanishing gradient problem of existing saliency maps. Experiments are conducted on five open datasets: FER2013, FERPlus, CASME, CASME II, and SAMM. To the best of our knowledge, by comparing with other methods offering source code (or pseudo code), we believe that our method would be the best on FERPlus, CASME, and CASME II and competitive on FER2013 and SAMM.

Track Everything: Limiting Prior Knowledge in Online Multi-Object Recognition.

This paper addresses the problem of online tracking and classification of multiple objects in an image sequence. Our proposed solution is to first track all objects in the scene without relying on object-specific prior knowledge, which in other systems can take the form of hand-crafted features or user-based track initialization. We then classify the tracked objects with a fast-learning image classifier, that is based on a shallow convolutional neural network architecture and demonstrate that object recognition improves when this is combined with object state information from the tracking algorithm. We argue that by transferring the use of prior knowledge from the detection and tracking stages to the classification stage, we can design a robust, general purpose object recognition system with the ability to detect and track a variety of object types. We describe our biologically inspired implementation, which adaptively learns the shape and motion of tracked objects, and apply it to the Neovision2 Tower benchmark data set, which contains multiple object types. An experimental evaluation demonstrates that our approach is competitive with the state-of-the-art video object recognition systems that do make use of object-specific prior knowledge in detection and tracking, while providing additional practical advantages by virtue of its generality.