A comparison of deep learning models for end-to-end face-based video retrieval in unconstrained videos

Face-based video retrieval (FBVR) is the task of retrieving videos that containing the same face shown in the query image. In this article, we present the first end-to-end FBVR pipeline that is able to operate on large datasets of unconstrained, multi-shot, multi-person videos. We adapt an existing audiovisual recognition dataset to the task of FBVR and use it to evaluate our proposed pipeline. We compare a number of deep learning models for shot detection, face detection, and face feature extraction as part of our pipeline on a validation dataset made of more than 4000 videos. We obtain 97.25% mean average precision on an independent test set, composed of more than 1000 videos. The pipeline is able to extract features from videos at $$\sim $$ ∼ 7 times the real-time speed, and it is able to perform a query on thousands of videos in less than 0.5 s.

COLOR IMAGE RETRIEVAL BASED ON FUZZY NEURAL NETWORK AND SWARM INTELLIGENCE TECHNIQUES

Image retrieval is an important system for retrieving similar images by searching and browsing in a large database. The image retrieval system can be a reliable tool for people to optimize the use of image accumulation, and finding efficient methods to retrieve images is very important. Recent decades have marked increased research interest in field image retrieval. To retrieve the images, an important set of features is used. In this work, a combination of methods was used to examine all the images and detect images in a database according to a query image. Linear Discriminant Analysis (LDA) was used for feature extraction of the images into the dataset. The images in the database were processed by extracting their important and robust features and storing them in the feature store. Likewise, the strong features were extracted for specific query images. By using some Meta Heuristic algorithms such as Cuckoo Search (CS), Ant Colony Optimization (ACO), and using an artificial neural network such as single-layer Perceptron Neural Network (PNN), similarity was evaluated. It also proposed a new two method by hybridized PNN and CS with fuzzy logic to produce a new method called Fuzzy Single Layer Perceptron Neural Network (FPNN), and Fuzzy Cuckoo Search to examine the similarity between features for query images and features for images in the database. The efficiency of the system methods was evaluated by calculating the precision recall value of the results. The proposed method of FCS outperformed other methods such as (PNN), (ACO), (CS), and (FPNN) in terms of precision and image recall. ABSTRAK: Imej dapatan semula adalah sistem penting bagi mendapatkan imej serupa melalui carian imej dan melayari pangkalan besar data. Sistem dapatan semula imej ini boleh dijadikan alat boleh percaya untuk orang mengoptimum penggunaan pengumpulan imej, dan kaedah pencarian yang berkesan bagi mendapatkan imej adalah sangat penting. Beberapa dekad yang lalu telah menunjukan banyak penyelidikan dalam bidang imej dapatan semula. Bagi mendapatkan imej-imej ini, ciri-ciri set penting telah digunakan. Kajian ini menggunakan beberapa kaedah bagi memeriksa semua imej dan mengesan imej dalam pangkalan data berdasarkan imej carian. Kami menggunakan Analisis Diskriminan Linear (LDA) bagi mengekstrak ciri imej ke dalam set data. Imej-imej dalam pangkalan data diproses dengan mengekstrak ciri-ciri penting dan berkesan daripadanya dan menyimpannya dalam simpanan ciri. Begitu juga, ciri-ciri penting ini diekstrak bagi imej carian tertentu. Persamaan dinilai melalui beberapa algoritma Meta Heuristik seperti Carian Cuckoo (CS), Pengoptimuman Koloni Semut (ACO), dan menggunakan lapisan tunggal rangkaian neural buatan seperti Rangkaian Neural Perseptron (PNN). Dua cadangan baru dengan kombinasi hibrid PNN dan CS bersama logik kabur bagi menghasilkan kaedah baru yang disebut Lapisan Tunggal Kabur Rangkaian Neural Perceptron (FPNN), dan Carian Cuckoo Kabur bagi mengkaji persamaan antara ciri carian imej dan imej pangkalan data. Nilai kecekapan kaedah sistem dinilai dengan mengira ketepatan mengingat pada dapatan hasil. Kaedah FCS yang dicadangkan ini mengatasi kaedah lain seperti (PNN), (ACO), (CS) dan (FPNN) dari segi ketepatan dan ingatan imej.

A New Content Based Search Mechanism for Image Retrieval Search Engine

In the growing world of technology, where everything is available in just one click, the user expectations has increased with time. In the era of Search Engines, where Google, Yahoo are providing the facility to search through text and voice and image , it has become a complex work to handle all the operations and lot more of data storage is needed. It is also a time consuming process. In the proposed Image retrieval Search Engine, the user enters the queried image and that image is being matched with the template images . The proposed approach takes the input image with 15% accuracy to 100% accuracy to retrieve the intended image by the user. But it is found that due to the efficiency of the applied algorithm, in all cases, the retrieved images are with the same accuracy irrespective of the input query image accuracy. This implementation is very much useful in the fields of forensic, defense and diagnostics system in medical field etc. .

Remote Network Monitoring Technology Based on Computer Technology

With the rapid development of Internet technology, using images to express the characteristics of things more direct, compared with text, audio, image expression content is more ambiguous, which makes the rapid increase of digital images on the Internet. Nowadays one of the hot directions of computer vision research is how to accurately and quickly retrieve the target image from a large amount of image data. This paper summarizes the development of image retrieval technology at home and abroad, and proposes an image search method based on color histogram and Chi-square distance. This paper discusses how to construct an image search system, which can search the image quickly, describe the color distribution of the photo with color histogram, divide the image into five regions, extract image features from the color histogram of each region, and then get the data set of multi-dimensional image features. Then the chi-square distance is used to calculate the similarity of color histogram, and the closest image is selected as the first similar image, which realizes the necessary logic of receiving query image and returning related results.

A Review on Content Based Image Retrieval System Features derived by Deep Learning Models

In a Content Based Image Retrieval (CBIR) System, the task is to retrieve similar images from a large database given a query image. The usual procedure is to extract some useful features from the query image, and retrieve images which have similar set of features. For this purpose, a suitable similarity measure is chosen, and images with high similarity scores are retrieved. Naturally the choice of these features play a very important role in the success of this system, and high level features are required to reduce the “semantic gap”. In this paper, we propose to use features derived from pre-trained network models from a deep- learning convolution network trained for a large image classification problem. This approach appears to produce vastly superior results for a variety of databases, and it outperforms many contemporary CBIR systems. We analyse the retrieval time of the method, and also propose a pre-clustering of the database based on the above-mentioned features which yields comparable results in a much shorter time in most of the cases. Keywords Content Based Image Retrieval Feature Selection Deep Learning Pre-trained Network Models Pre-clustering

TAE-Net: Task-Adaptive Embedding Network for Few-Shot Remote Sensing Scene Classification

Recently, approaches based on deep learning are quite prevalent in the area of remote sensing scene classification. Though significant success has been achieved, these approaches are still subject to an excess of parameters and extremely dependent on a large quantity of labeled data. In this study, few-shot learning is used for remote sensing scene classification tasks. The goal of few-shot learning is to recognize unseen scene categories given extremely limited labeled samples. For this purpose, a novel task-adaptive embedding network is proposed to facilitate few-shot scene classification of remote sensing images, referred to as TAE-Net. A feature encoder is first trained on the base set to learn embedding features of input images in the pre-training phase. Then in the meta-training phase, a new task-adaptive attention module is designed to yield the task-specific attention, which can adaptively select informative embedding features among the whole task. In the end, in the meta-testing phase, the query image derived from the novel set is predicted by the meta-trained model with limited support images. Extensive experiments are carried out on three public remote sensing scene datasets: UC Merced, WHU-RS19, and NWPU-RESISC45. The experimental results illustrate that our proposed TAE-Net achieves new state-of-the-art performance for few-shot remote sensing scene classification.

Image-Based Hot Pepper Disease and Pest Diagnosis Using Transfer Learning and Fine-Tuning

Past studies of plant disease and pest recognition used classification methods that presented a singular recognition result to the user. Unfortunately, incorrect recognition results may be output, which may lead to further crop damage. To address this issue, there is a need for a system that suggest several candidate results and allow the user to make the final decision. In this study, we propose a method for diagnosing plant diseases and identifying pests using deep features based on transfer learning. To extract deep features, we employ pre-trained VGG and ResNet 50 architectures based on the ImageNet dataset, and output disease and pest images similar to a query image via a k-nearest-neighbor algorithm. In this study, we use a total of 23,868 images of 19 types of hot-pepper diseases and pests, for which, the proposed model achieves accuracies of 96.02 and 99.61%, respectively. We also measure the effects of fine-tuning and distance metrics. The results show that the use of fine-tuning-based deep features increases accuracy by approximately 0.7–7.38%, and the Bray–Curtis distance achieves an accuracy of approximately 0.65–1.51% higher than the Euclidean distance.

IMAGE RETRIEVAL BASED ON DISCRETE CURVELET TRANSFORM

Content-Based Image Retrieval (CBIR) is a process of searching for an image according to the content or feature that is within it. Nowadays, most image retrieval applications have been developed to meet these needs, so this application will provide comfort in introducing and searching for an image. This paper proposed a standard structured framework with three stages: Preprocessing is the first step, in which noise from images is removed using various filters. The filters' results are compared to determine the best and most appropriate filter for the images. Feature Extraction of images using Curvelet Transform is the second stage. The third stage includes similarity measurement between query image features to database image features and extracting the identical image from the image dataset. The system was performed using Matlab 2017b, GUI and, with ten different classes of 1000 images using a coral database. The results show improved performance of precision and recall when higher decomposition levels are used.

Image retrieval based on swarm intelligence

To keep pace with the development of modern technology in this information technology era, and the immense image databases, whether personal or commercial, are increasing, is requiring the management of these databases to strong and accurate systems to retrieve images with high efficiency. Because of the swarm intelligence algorithms are great importance in solving difficult problems and obtaining the best solutions. Here in this research, a proposed system is designed to retrieve color images based on swarm intelligence algorithms. Where the algorithm of the ant colony optimization (ACOM) and the intelligent water drop (IWDM) was used to improve the system's work by conducting the clustering process in these two methods on the features extracted by annular color moment method (ACM) to obtain clustered data, the amount of similarity between them and the query image, is calculated to retrieve images from the database, efficiently and in a short time. In addition, improving the work of these two methods by hybridizing them with fuzzy method, fuzzy gath geva clustering algorithm (FGCA) and obtaining two new high efficiency hybrid algorithms fuzzy ant colony optimization method (FACOM) and fuzzy intelligent water drop method (FIWDM) by retrieving images whose performance values are calculated by calculating the values of precision, recall and the f-measure. It proved its efficiency by comparing it with fuzzy method, FGCA and by methods of swarm intelligence without hybridization, and its work was excellent.

Street-Level Image Localization Based on Building-Aware Features via Patch-Region Retrieval under Metropolitan-Scale

The aim of image-based localization (IBL) is to localize the real location of query image by matching reference image in database with GNSS-tags. Popular methods related to IBL commonly use street-level images, which have high value in practical application. Using street-level image to tackle IBL task has the primary challenges: existing works have not made targeted optimization for urban IBL tasks. Besides, the matching result is over-reliant on the quality of image features. Methods should address their practicality and robustness in engineering application, under metropolitan-scale. In response to these, this paper made following contributions: firstly, given the critical of buildings in distinguishing urban scenes, we contribute a feature called Building-Aware Feature (BAF). Secondly, in view of negative influence of complex urban scenes in retrieval process, we propose a retrieval method called Patch-Region Retrieval (PRR). To prove the effectiveness of BAF and PRR, we established an image-based localization experimental framework. Experiments prove that BAF can retain the feature points that fall on the building, and selectively lessen the feature points that fall on other things. While this effectively compresses the storage amount of feature index, we can also improve recall of localization results; implemented in the stage of geometric verification, PRR compares matching results of regional features and selects the best ranking as final result. PRR can enhance effectiveness of patch-regional feature. In addition, we fully confirmed the superiority of our proposed methods through a metropolitan-scale street-level image dataset.