Convolutional neural networks (CNNs) have become a classic approach to solving challenging computer vision problems. Much of its success is due to its ability to discover optimal filters that capture non-trivial spatial relationships in data. Other vital components include advances in optimization, regularization, and overfitting prevention strategies. However, recently, researchers have observed closely the connections between what CNNs learn in the layers that capture low-level features and filter-banks such as Gabor filters. Gabor filters have been used in computer vision tasks long before CNNs were popularized with good performance. This paper presents a review of the literature concerning approaches that involve both Gabor filters and CNNs. We pay close attention to successes and opportunities for future research in the intersection of these two computer vision tools.

A Review of Convolutional Neural Networks and Gabor Filters in Object Recognition

Vehicle make and model classification is crucial to the operation of an intelligent transportation system (ITS). Fine-grained vehicle information such as make and model can help officers uncover cases of traffic violations when license plate information cannot be obtained. Various techniques have been developed to perform vehicle make and model classification. However, it is very hard to identify the make and model of vehicles with highly similar visual appearances. The classifier contains a lot of potential for mistakes because the vehicles look very similar but have different models and manufacturers. To solve this problem, a fine-grained classifier based on convolutional neural networks with a multi-task learning approach is proposed in this paper. The proposed method takes a vehicle image as input and extracts features using the VGG-16 architecture. The extracted features will then be sent to two different branches, with one branch being used to classify the vehicle model and the other to classify the vehicle make. The performance of the proposed method was evaluated using the InaV-Dash dataset, which contains an Indonesian vehicle model with a highly similar visual appearance. The experimental results show that the proposed method achieves 98.73% accuracy for vehicle make and 97.69% accuracy for vehicle model. Our study also demonstrates that the proposed method is able to improve the performance of the baseline method on highly similar vehicle classification problems.

https://www.mdpi.com/2313-433X/8/11/293/pdf?version=1666429693

CNN-Based Classification for Highly Similar Vehicle Model Using Multi-Task Learning

Deep convolutional neural networks have been widely used for saliency detection. However, most of the previous works focus on the visible light image. In this paper, there are mainly two contributions. First, we propose a new architecture named Multi-level Up-sampling Network (MLUNet) for infrared (IR) ship object saliency detection. Specifically, the architecture of MLUNet is an Encoder-Decoder like network embedded with subtraction feature filtering module (SFFM). The encoder uses the DenseNet like architecture, and the decoder part use two upsampling methods, which are deconvolution and sub-pixel convolution. SFFM is a feature subtraction module which is in charge of feature filtering. In our proposed MLUNet, SFFM is embedded after each convolution and deconvolution block. Secondly, we construct an IR ship object image dataset for saliency detection. This dataset includes 3845 IR images and ground-truth images with different backgrounds and different objects. Experimental results show that our method outperforms the state-of-the-art methods in terms of regional evaluation measures.

Multi-level Up-sampling Network for Infrared Ship Saliency Object Detection

AbstractIn recent years, for recognizing sign language, several hardware approaches have been developed using the leap motion controller and Kinect sensors. The sensor-based approaches were costly, and the complexity of designing was high. In machine learning approaches, it was found that less accuracy of prediction occurs as compared to other approaches. As a solution, machine learning-based approaches using image processing evolved as an approach with better prediction and accuracy. In this paper, deep learning approach for sign language recognition (SLR) system has been proposed using AlexNet model. AlexNet is pretrained model, and it is being tested on Indian Sign Language (ISL) Dataset pf Stanford University comprising of alphabets (A–Z) and numerals (0–9). The proposed model has yielded an accuracy of 99.6%.KeywordsAlexNetDeep learningSLRRectified linear unit layer 

AlexNet Model for Sign Language Recognition

It is common to find objects that resemble animals on purpose (e.g., toys). While the perception of such objects as animal-like seems obvious to humans, such “Animal bias” for zoomorphic objects turned out to be a striking discrepancy between the human brain and artificial visual systems known as deep neural networks (DNNs). We provide a computational understanding of the human Animal bias. We successfully induced this bias in DNNs trained explicitly with zoomorphic objects. Alternative training schedules, focusing on previously identified differences between the brain and DNNs, failed to cause an Animal bias. Specifically, we considered the superordinate distinction between animate and inanimate classes, the sensitivity for faces and bodies, the bias for shape over texture, and the role of ecologically valid categories. These findings provide computational support that the Animal bias for zoomorphic objects is a unique property of human perception yet can be explained by human learning history.

A computational understanding of zoomorphic perception in the human brain

This paper evaluates two deep learning techniques that are basic Convolutional Neural Network (CNN) and AlexNet along with a classical local descriptor that is Bag of Features (BoF) with Speeded-Up Robust Feature (SURF) and Support Vector Machine (SVM) classifier for indoor object recognition. A publicly available dataset, MCIndoor20000, has been used in this experiment that consists of doors, signage, and stairs images of Marshfield Clinic. Experimental results indicate that AlexNet achieves the highest accuracy followed by basic CNN and BoF. Furthermore, the results also show that BoF, a machine learning technique, can also produce a high accuracy performance as basic CNN, a deep learning technique, for image recognition.

/pdf/evaluation-of-basic-convolutional-neural-network-alexnet-and-49ob3nir6e.pdf

Fatin Amira Fammy

Papers

Evaluation of Basic Convolutional Neural Network, AlexNet and Bag of Features for Indoor Object Recognition