Solving Image Classification Problems Research Articles

Simulating an Integrated Photonic Image Classifier for Diffractive Neural Networks.

The slowdown of Moore's law and the existence of the "von Neumann bottleneck" has led to electronic-based computing systems under von Neumann's architecture being unable to meet the fast-growing demand for artificial intelligence computing. However, all-optical diffractive neural networks provide a possible solution to this challenge. They can outperform conventional silicon-based electronic neural networks due to the significantly higher speed of the propagation of optical signals (≈108 m.s-1) compared to electrical signals (≈105 m.s-1), their parallelism in nature, and their low power consumption. The integrated diffractive deep neural network (ID2NN) uses an on-chip fully passive photonic approach to achieve the functionality of neural networks (matrix-vector operations) and can be fabricated via the CMOS process, which is technologically more amenable to implementing an artificial intelligence processor. In this paper, we present a detailed design framework for the integrated diffractive deep neural network and corresponding silicon-on-insulator integration implementation through Python-based simulations. The performance of our proposed ID2NN was evaluated by solving image classification problems using the MNIST dataset.

Sampling based spherical transformer for 360 degree image classification

Using convolutional neural networks for 360° images can induce sub-optimal performance due to distortions entailed by a planar projection. The distortion gets deteriorated when a rotation is applied to the 360° image. Thus, many researches based on convolutions attempt to reduce the distortions to learn accurate representation. In contrast, we leverage the transformer architecture to solve image classification problems for 360° images. Using the proposed transformer for 360° images has two advantages. First, our method does not require the erroneous planar projection process by sampling pixels from the sphere surface. Second, our sampling method based on regular polyhedrons makes low rotation equivariance errors, because specific rotations can be reduced to permutations of faces. In experiments, we validate our network on two aspects, as follows. First, we show that using a transformer with highly uniform sampling methods can help reduce the distortion. Second, we demonstrate that the transformer architecture can achieve rotation equivariance on specific rotations. We compare our method to other state-of-the-art algorithms using the SPH-MNIST, SPH-CIFAR, and SUN360 datasets and show that our method is competitive with other methods.

DenseNet Based Model for Plant Diseases Diagnosis

The biggest threat to the safety of food is plant diseases. They have the ability to dramatically lower the quantity and quality of agricultural products. Recognizing plant diseases is the biggest issue in the agricultural industries. Convolutional Neural Networks (CNN) are effective in solving image classification problems in computer vision. Numerous deep learning architectures have been used to diagnose plant diseases. This study presents a transfer learning-based model for identifying diseases in plant leaves. In this paper, a CNN classifier based on transfer learning model called DenseNet201 are proposed. An analysis of four deep learning models (VGG16, Inception V3, ResNet152V2, and DenseNet201) done to see which one can detect plant diseases with the greatest degree of accuracy. Web based application developed for plant disease diagnosing from defected leaf image and the proposed model which identify the disease and give the recommended treatment. The used images dataset contains 28310 leaves photos of 3 crops, tomato, potato and pepper divided into 15 different classes, 9 disorders and one healthy class for tomato, 2 disorders and one healthy class for potato and 1 disorder and one healthy for pepper. In our experimental, the results shows that the proposed model achieves the highest training accuracy of 99.44% and validation accuracy of 98.70%.

Unsupervised segmentation of microstructural images of steel using data mining methods

A novel and efficient data mining method for the segmentation of microstructural images of low-carbon steel is presented. Microstructural characterization has been the focus of many works in the field of materials science because microstructure is the fundamental element in understanding the link between process and property. Recently, deep-learning-based methods have been actively employed for microstructural classification since it has shown outstanding performance for solving image classification problems. However, previous applications of deep learning models to microstructural classification revealed limitations in that not only do they require the time-consuming labeling process but it is also still difficult to obtain a satisfactory result, especially for steel microstructures containing substances developed by displacive transformation. In this study, we propose a rule-based segmentation method that not only works without labeled images but also requires no prior knowledge of the number of microstructural constituents in each image. This unsupervised inference algorithm captures the morphological features of each microstructure and automatically finds the optimal number of microstructures having similar characteristics using a Bayesian Gaussian mixture model. The viability of our method is demonstrated by qualitative and quantitative evaluations with optical microscopy images of steel composed of different microstructures taken under different imaging conditions.

Comparative Study on Reliability of Transfer Learning to Classify Plant-Based Diseases

One of the essential factors contributing to a plant's growth is identifying and preventing diseases in the early stages. Healthy plants are essential for a rich production. Recent advances in Deep learning - a subset of Artificial Intelligence and Machine Learning are playing a pivotal role in solving image classification problems and can be applied to the agricultural sector for crop surveillance and early anomaly identification. For this research, we used an open-source dataset of leaf images divided into three classes, two of which are the most common disease types found on many crops; the graphical characterizations for the three classes are images of leaves with Powdery Residue, images of leaves with Rusty Spots, and images of Healthy leaves. The primary objective of this research is to present a pre-trained ImageNet network architecture that is well suited for dealing with plant-based data, even when sample sizes collected are limited. We used different convolutional neural network-based architectures such as InceptionV3, MobileNetV2, Xception, VGG16, and VGG19 to classify plant leaf images with visually different representations of each disease. Xception, MobileNetV2, and DenseNet had a considerable advantage over all the performance metrics recorded among the other networks trained.

Real UAV-Bird Image Classification Using CNN with a Synthetic Dataset

A large amount of training image data is required for solving image classification problems using deep learning (DL) networks. In this study, we aimed to train DL networks with synthetic images generated by using a game engine and determine the effects of the networks on performance when solving real-image classification problems. The study presents the results of using corner detection and nearest three-point selection (CDNTS) layers to classify bird and rotary-wing unmanned aerial vehicle (RW-UAV) images, provides a comprehensive comparison of two different experimental setups, and emphasizes the significant improvements in the performance in deep learning-based networks due to the inclusion of a CDNTS layer. Experiment 1 corresponds to training the commonly used deep learning-based networks with synthetic data and an image classification test on real data. Experiment 2 corresponds to training the CDNTS layer and commonly used deep learning-based networks with synthetic data and an image classification test on real data. In experiment 1, the best area under the curve (AUC) value for the image classification test accuracy was measured as 72%. In experiment 2, using the CDNTS layer, the AUC value for the image classification test accuracy was measured as 88.9%. A total of 432 different combinations of trainings were investigated in the experimental setups. The experiments were trained with various DL networks using four different optimizers by considering all combinations of batch size, learning rate, and dropout hyperparameters. The test accuracy AUC values for networks in experiment 1 ranged from 55% to 74%, whereas the test accuracy AUC values in experiment 2 networks with a CDNTS layer ranged from 76% to 89.9%. It was observed that the CDNTS layer has considerable effects on the image classification accuracy performance of deep learning-based networks. AUC, F-score, and test accuracy measures were used to validate the success of the networks.

Study of the Effect of Combining Activation Functions in a Convolutional Neural Network

Convolutional Neural Networks (CNNs) have proven to be an effective approach for solving image classification problems. The output, the accuracy and the computational efficiency of a CNN are determined mainly by the architecture, the convolutional filters, and the activation functions. Based on the importance of an activation function, in this paper, nine new activation functions based on combinations of classical functions such as ReLU and sigmoid are presented. Also, a study about the effects caused by the activation functions in the performance of a CNN is presented. First, every new function is described, also, their graphs, analytic forms and derivatives are presented. Then, a traditional CNN model with each new activation function is used to classify three 10-class databases: MNIST, Fashion MNIST and a handwritten digit database created by us. Experimental results illustrate that some of the proposed activation functions lead to better performances on classifying than classical activation functions. Moreover, our study demonstrated that the accuracy of a CNN could be increased by 1.18% with the new proposed activation functions.

Optimizing Spatiotemporal Feature Learning in 3D Convolutional Neural Networks With Pooling Blocks

Image data contain spatial information only, thus making two-dimensional (2D) Convolutional Neural Networks (CNN) ideal for solving image classification problems. On the other hand, video data contain both spatial and temporal information that must be simultaneously analyzed to solve action recognition problems. 3D CNNs are successfully used for these tasks, but they suffer from their extensive inherent parameter set. Increasing the network’s depth, as is common among 2D CNNs, and hence increasing the number of trainable parameters does not provide a good trade-off between accuracy and complexity of the 3D CNN. In this work, we propose Pooling Block (PB) as an enhanced pooling operation for optimizing action recognition by 3D CNNs. PB comprises three kernels of different sizes. The three kernels simultaneously sub-sample feature maps, and the outputs are concatenated into a single output vector. We compare our approach with three benchmark 3D CNNs (C3D, I3D, and Asymmetric 3D CNN) and three datasets (HMDB51, UCF101, and Kinetics 400). Our PB method yields significant improvement in 3D CNN performance with a comparatively small increase in the number of trainable parameters. We further investigate (1) the effect of video frame dimension and (2) the effect of the number of video frames on the performance of 3D CNNs using C3D as the benchmark.

Learning Diatoms Classification from a Dry Test Slide by Holographic Microscopy.

Diatoms are among the dominant phytoplankters in marine and freshwater habitats, and important biomarkers of water quality, making their identification and classification one of the current challenges for environmental monitoring. To date, taxonomy of the species populating a water column is still conducted by marine biologists on the basis of their own experience. On the other hand, deep learning is recognized as the elective technique for solving image classification problems. However, a large amount of training data is usually needed, thus requiring the synthetic enlargement of the dataset through data augmentation. In the case of microalgae, the large variety of species that populate the marine environments makes it arduous to perform an exhaustive training that considers all the possible classes. However, commercial test slides containing one diatom element per class fixed in between two glasses are available on the market. These are usually prepared by expert diatomists for taxonomy purposes, thus constituting libraries of the populations that can be found in oceans. Here we show that such test slides are very useful for training accurate deep Convolutional Neural Networks (CNNs). We demonstrate the successful classification of diatoms based on a proper CNNs ensemble and a fully augmented dataset, i.e., creation starting from one single image per class available from a commercial glass slide containing 50 fixed species in a dry setting. This approach avoids the time-consuming steps of water sampling and labeling by skilled marine biologists. To accomplish this goal, we exploit the holographic imaging modality, which permits the accessing of a quantitative phase-contrast maps and a posteriori flexible refocusing due to its intrinsic 3D imaging capability. The network model is then validated by using holographic recordings of live diatoms imaged in water samples i.e., in their natural wet environmental condition.

Biomedical image classification made easier thanks to transfer and semi-supervised learning

Background and objectivesDeep learning techniques are the state-of-the-art approach to solve image classification problems in biomedicine; however, they require the acquisition and annotation of a considerable volume of images. In addition, using deep learning libraries and tuning the hyperparameters of the networks trained with them might be challenging for several users. These drawbacks prevent the adoption of these techniques outside the machine-learning community. In this work, we present an Automated Machine Learning (AutoML) method to deal with these problems. MethodsOur AutoML method combines transfer learning with a new semi-supervised learning procedure to train models when few annotated images are available. In order to facilitate the dissemination of our method, we have implemented it as an open-source tool called ATLASS. Finally, we have evaluated our method with two benchmarks of biomedical image classification datasets. ResultsOur method has been thoroughly tested both with small datasets and partially annotated biomedical datasets; and, it outperforms, both in terms of speed and accuracy, the existing AutoML tools when working with small datasets; and, might improve the accuracy of models up to a 10% when working with partially annotated datasets. ConclusionsThe work presented in this paper allows the use of deep learning techniques to solve an image classification problem with few resources. Namely, it is possible to train deep models with small, and partially annotated datasets of images. In addition, we have proven that our AutoML method outperforms other AutoML tools both in terms of accuracy and speed when working with small datasets.

Extreme learning machine with coefficient weighting and trained local receptive fields for image classification

Local receptive fields based extreme learning machine (ELM-LRF) is widely used to solve image classification problems. However, the performance of ELM-LRF is limited by the single generation method of local receptive fields and the simple network structure. In order to solve these problems and make full use of image information to improve classification accuracy, extreme learning machine with coefficient weighting and trained local receptive fields (ELM-WLRF) is proposed based on ELM-LRF. The structure is mainly composed of convolution blocks, weighting blocks, dimensionality reduction and classification layers. In the convolution block, the principle of the ELM and the method of grouping calculation are used to train the local receptive fields of the two convolutional layers. The trained local receptive fields are used to extract identifiable feature information in the image more stably and adequately. In the weighting block, the principles of ELM and ELM autoencoder (ELM-AE) are used to train channel and spatial weighting coefficients to improve the recognizability of features. In the dimensionality reduction and classification layers, the approximate empirical kernel map (EKM) is used to train the connection weight matrix between each layer to further improve the network training speed and classification accuracy. To demonstrate the effectiveness of the proposed method, ELM-WLRF is tested on the MNIST, NORB and CIFAR-10 databases. The experimental results show that ELM-WLRF achieves superior classification accuracy, i.e. 99.27%, 98.03% and 60.14% respectively, and requires shorter training time compared with other state-of-the-art ELM-LRF-based algorithms.

Probability Voting-based Ensemble of Convolutional Neural nets Classifiers for Image Classification

This study explores an ensemble technique for building a composite of pre-trained VGG16, VGG19, and Resnet56 classifiers using probability voting-based technique. The resulted composite classifiers were tested to solve image classification problems using a subset of Cifar10 dataset. The classifier performance was measured using accuracy metric. Some experimentation results show that the ensemble methods of pre-trained VGG19-Resnet56 and VGG16-VGG19-Resnet models outperform the accuracy of its individual model and other composite models made of these three models.

Visual word coding based on difference maximization

Image classification is an important topic in computer vision, which becomes more and more challenging due to the rapid increase of the amounts and categories of images, as well as the different geometric deformations and illumination variations existing in image objects. “Bag-of-Features” model, also known as “Bag-of-Words” model, plays a fundamental and crucial role in generating efficient and discriminative image content representations by using local descriptors such as visual words, making it widely used in solving image classification problems. However, because of the weak discriminative power and strong ambiguity of the low-level visual features, the visual codebook, i.e., a set of visual words, generated in this model is usually over-completed and inconsistent for capturing image semantics. To address this issue, we propose a novel visual word coding algorithm in this paper based on difference maximization technique to improve the generated codebook model. Instead of mapping an image feature vector to one or multiple nearest visual words, the proposed approach utilizes a group of the nearest and the farthest visual words together in the coding process. Consequently, the representative variations of different image features are well kept and strengthened, which can then improve the discriminative power of the visual word descriptor significantly. We examine the performance of our visual word coding model extensively on four standard real-world image datasets, demonstrating that it captures image semantic content more accurately and achieves superior classification performance.