Object Tracking Via ImageNet Classification Scores

Context-aware long-term correlation tracking with hierarchical convolutional features

Multiple correlation filters with gaussian constraint for fast online tracking

Handwritten digit recognition, a task in computer vision, is critical for applications such as postal automation, banking, and digitization of forms. Traditional approaches have leveraged statistical models, but the rise of deep learning, particularly Convolutional Neural Networks (CNN) and Long-Short Term Memory (LSTM), has revolutionized the field. However, a comprehensive performance comparison of CNN and LSTM architectures in the context of handwritten digit classification remains underexplored. This study aimed to address this gap by evaluating and comparing CNN and LSTM models for the classification of handwritten digits. Two machine learning models – Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) were trained with the preprocessed handwritten digit dataset. The CNN model was designed with multiple convolutional and pooling layers, along with dropout for regularization. The LSTM model was designed with LSTM layers to capture sequential patterns in the data, followed by a dense layer for classification. The models were implemented in python, evaluated and compared based on accuracy, precision, recall and F1-score. The evaluation and comparison results indicate that CNN achieved 99.31% accuracy, 99.0% precision, 99.0% recall, and a 99.0% F1-score, while LSTM achieved 98.90% accuracy, 99.0% precision, 99.0% recall, and a 99.0% F1-score. The results demonstrated that CNN outperformed LSTM in terms of accuracy and misclassification errors, making it the optimal choice for image-based handwritten digit recognition. This finding underscores the efficiency of CNN in addressing challenges related to digit recognition, contributing to the advancement of automated digit classification systems and improving the accuracy of image-based classification tasks.

CFGVF: An improved correlation filters based visual tracking algorithm

Object tracking is one of the most important tasks in computer vision, many researchers have proposed a lot of effective methods to solve the all kinds of the recent problems in object tracking. During the movement of the targets, the change of object scale may produce a lot of difficulties in object tracing problems, and it may lead to be failed in tracking. The paper proposed method is designed both correlation filter and particle filter. In order to achieve the purpose of tracking, the process of correlation filter is to learn an array of filters so that the response value obtained by the convolution of the learned filters is the largest. The particle filter guides the sample particles to the distribution mode of the target state. The samples of the particle strategy can effectively deal with the large scale variations of the target during movement. Experiment on the OTB database shows that the paper proposed method makes good performance.

Image classification is the elementary and widely addressed task in computer vision. Classification of monument architecture is a real challenge due to its structural complexity. Analyzing architecture using the machine learning or data mining techniques can be useful. Various feature descriptors are used for large scale image classification. In this context, optimal feature descriptors as filters are used in Convolutional Neural Network. When analyzing architecture using the machine learning or data mining techniques, it can be useful to infer whether it is Cathedral or Indian Mughal Monuments in architecture involved. This paper focuses on architecture recognition, which has been studied extensively in the history literature, using Convolutional Neural Network (CNN) based on highly effective Tensorflow, an open source software library. The important task is to generalize the model to perform well on the new database. Ontology performs a significance role in learning to categorize given picture into metaphysical classes. A deep learning based approach is used to train the convolutional neural network. In the proposed work, images are classified into Cathedrals and Indian Mughal Monuments. Experimental results demonstrate that the method can successfully classify if the given image is Cathedral or Indian Mughal Monument which mainly includes the Taj Mahal and Char Minar. Proposed method substantially outperform state-of art approaches which is demonstrated by satisfactory result obtained on the dataset.

Object classification has been one of the main tasks in computer vision. With the fast development of deep learning, its performance in image classification and object recognition has presented dramatic improvements. In this paper, we propose a new deep convolutional neural network (CNN) architecture for robust object classification. The proposed model is fused with three traditional CNN approaches, Densenet201, Resnet50, and our proposed residual CNN. The fused network architecture allows parallel processing of the multiple networks for keeping the system sped up. A single shot deep convolution network is trained as an object detector to generate all possible candidates of different object classes. The output of each neural network is representing a single vote that is used in the classification process. 3-to-1 voting criteria are applied in the final classification decision between the candidate object classes. Several experiments were conducted to evaluate the performance of the proposed network. The experimental results show that the proposed approach has better performance than the networks used in the fusion process when they act individually. It also has lower miss rates when compared to several states of art methodologies.

In recent years, the convolutional neural network (CNN) plays a vital role in hyperspectral image classification and performs more competitively than many other methods. However, in order to pursue better performance, most of existing CNN-based methods just simply stack rather deep convolutional layers. Although they improve the classification accuracy to a certain extent, they result in plenty of network parameters. In this paper, a light-weighted directionally separable dilated CNN with hierarchical attention feature fusion (DSD-HAFF) is proposed to solve these problems. First, two global dense dilated CNN branches that focus on two spatial directions separately are constructed to extract and reuse spatial information as much as possible. Second, a hierarchical attention feature fusion branch that consists of several coordinate attention blocks (CABs) is constructed. Hierarchical features from two directionally separable dilated CNN branches are adopted as inputs of CABs. In this way, the structure can not only fully incorporate hierarchical features, but also significantly reduce the network parameters. Meanwhile, the hierarchical attention feature fusion branch incorporates features from high-level to low-level in the kernel-number pyramid strategy. Experimental results on three popular benchmark datasets demonstrate that the DSD-HAFF achieves better performance and has a much smaller number of network parameters than the other state-of-the-art methods.

The performance of image classification technology based on deep network has been greatly improved, making computer vision enter the stage of industrialization and be gradually applied to many aspects of human work and life. As a typical classification task in computer vision, human behavior recognition has immeasurable potential value in medical, family, transportation, and other scenarios. At the same time, in the field of competitive sports, the integration of artificial intelligence technology and sports technical and tactical analysis is undoubtedly an important way to innovate and improve the technical and tactical level. Taking karate as an example, the study of athletes’ training and competition videos is an important means and method for technical and tactical analysis in competitive sports. Traditional tactical intelligence analysis methods have many shortcomings, such as high labor cost, serious data loss, long delay, and low accuracy. Therefore, based on the convolutional neural network, this paper establishes a new graph convolution model for automatic intelligent analysis of karate athletes’ technical action recognition, action frequency statistics, and trajectory tracking. The technology effectively makes up for the disadvantages of traditional tactical intelligence analysis methods. The research results show that the new topology map construction method has a significant effect on improving the accuracy of behavior recognition and also lays a foundation for technical and tactical analysis.

A large portion of India's population relies primarily on agriculture for their livelihood. Farmers suffer a considerable amount of loss due to the innumerable diseases affecting their plants. Detection of such plant diseases with the human eye often yields inaccurate results. Furthermore, to correctly identify the disease, the individual assessing the plant should be an expert in the respective field. The diagnosis of plant illness is a visual task and thus, many computer vision techniques have been used previously for tackling it. Recently, convolutional Neural Networks have shown excellent results in many computer vision tasks. This study develops an application for plant disease classification by comparing the results obtained by training two convolutional neural networks, one from scratch and one by the transfer learning method. Both achieved a validation accuracy of 86 percent and 96 percent, respectively. The system was developed in the form of a web application for both mobile and web devices using the model, which is capable of functioning without any network requirements.

Owing to the wide spread of the 3D objects technologies, learning 3D objects labeling and segmentation is becoming one of the most provocative tasks in computer vision and digital multimedia. Recently, convolutional neural network (CNN) has proved itself as a powerful model in segmentation and classification by giving excellent performances. The use of a graphical model such as conditional random field (CRF) contributes further in capturing contextual information and thus improving the segmentation performance. In this paper, we implement a deep architecture in order to segment and label 3D shape parts by considering both spectral and geometric features via a combined framework consisting of a CNN and CRF models. First, low-level features are used to learn deep features using a CNN model, and then formulate the deep CRF model with a CNN-based unary and pairwise potential functions to effectively extract the semantic correlations between adjacent triangles on the mesh. By comparing our results with those from several state-of-the-art, our method shows promising potentials.

Background of study: Food image classification is a challenging task in computer vision, particularly when dealing with traditional food items that exhibit subtle visual variations. While Convolutional Neural Networks (CNNs) have long been the standard for image recognition, their limitation in capturing long-range dependencies has led to the emergence of Vision Transformers (ViTs). In this context, the classification of Indonesian traditional cakes offers a culturally rich yet complex problem for automated image recognition systems.Aims and scope of paper: This study aims to conduct a comparative analysis between EfficientNet-B0 (CNN-based) and ViT-B/16 (Transformer-based) architectures in classifying eight categories of Indonesian traditional cakes. The research evaluates not only classification accuracy but also the strengths and limitations of each model in handling fine-grained visual distinctions.Methods: Both models were fine-tuned using the “Kue Indonesia” dataset from Kaggle. The methodology includes image preprocessing, model training with consistent parameters, and evaluation using accuracy, precision, recall, and F1-score. A confusion matrix was also used to visualize misclassifications and analyze per-class performance.Result: ViT-B/16 achieved slightly higher accuracy (96.25%) compared to EfficientNet-B0 (95.62%). ViT performed better in classes with subtle variations, such as kue lapis and kue dadar gulung, while EfficientNet-B0 showed superior efficiency and high accuracy on visually distinct cakes.Conclusion: Both CNN and ViT models demonstrate strong performance in traditional food classification. ViT is more robust in fine-grained visual analysis, whereas EfficientNet-B0 is preferable for resource-constrained environments. This study highlights the role of AI in supporting digital preservation of culinary heritage.

Brain cancer is a life-threatening disease that affects the brain. Detecting a tumor at an early stage becomes critical to saving lives. For the study of the brain and various medical images, magnetic resonance imaging and image segmentation algorithms have grown to be important medical diagnostic tools. Object detection and classification are important tasks in computer vision and image processing. Medicinal images play an important part in the diagnosis of tumors, but the detection of cancer faces many challenges. In this paper, the developed model has been suggested to classify CT-scan images of the brain at the beginning of production to have a tumor or recognition. It is built on a Deep Convolutional Neural Network (DCNN), and it has a particular architecture for detecting complex data features. Python has used to build the proposed model. The total dataset of the proposed network contains 4,500 images. The system used 3,600 images for training and 900 images for testing. The total number of parameters during the learning of the proposed model is 2,398,750. The highest accuracy is reached when the image size is 50x50 pixels and the number of epochs is 100. The suggested network has high performance to predict the tumor in different conditions. The work's experimental outcomes demonstrated the tumor's effectiveness with 98.5 % accuracy with in a very limited time. This model supports healthcare practitioners in making decisions about the presence or absence of disease. It is a more accurate, fast, and effective method to recognize the brain cancer based on Deep Learning technology.

Fast object tracking system is important to Augmented Reality (AR). However, tracking multiple objects fast and accurately is not a trivial task in computer vision, especially when the objects become crowded and frequently occluded. In this paper, we design an efficient segmentation model along with the tracking and carefully study the contributions of pixel-level appearance modeling in the presence of occlusions. Different from previous joint tracking and segmentation methods that depend on bounding boxes for instance segmentation, we proposed an anchor free model to speed the object mask estimation. Specifically, our proposed approach jointly addresses instance segmentation and tracking with a single convolutional network for the first time, and generates object masks to explicitly separate overlapping objects in tracking. As pointed out by several previous works, the goals of tracking and segmentation are not always consistent, we deliberately design three homogeneous branches in our network to balance the competitive three functional parts of objectness, masks, and re-identification simultaneously and constrain them in a comparable scale. With the effective three functional parts, our method could run fast for the time-sensitive tracking task and meanwhile relieve the effects of appearance noise brought by the occlusions. Experimental results on the MOTS benchmark demonstrate the effectiveness of the proposed method.

Deep learning has already been proved as a powerful state-of-the-art technique for many image understanding tasks in computer vision and other applications including remote sensing (RS) image analysis. Unmanned aircraft systems (UASs) offer a viable and economical alternative to a conventional sensor and platform for acquiring high spatial and high temporal resolution data with high operational flexibility. Coastal wetlands are among some of the most challenging and complex ecosystems for land cover prediction and mapping tasks because land cover targets often show high intra-class and low inter-class variances. In recent years, several deep convolutional neural network (CNN) architectures have been proposed for pixel-wise image labeling, commonly called semantic image segmentation. In this paper, some of the more recent deep CNN architectures proposed for semantic image segmentation are reviewed, and each model’s training efficiency and classification performance are evaluated by training it on a limited labeled image set. Training samples are provided using the hyper-spatial resolution UAS imagery over a wetland area and the required ground truth images are prepared by manual image labeling. Experimental results demonstrate that deep CNNs have a great potential for accurate land cover prediction task using UAS hyper-spatial resolution images. Some simple deep learning architectures perform comparable or even better than complex and very deep architectures with remarkably fewer training epochs. This performance is especially valuable when limited training samples are available, which is a common case in most RS applications.