Search

Sketch-based image retrieval (SBIR) utilizes sketches to search for images containing similar objects or scenes. Due to the proliferation of touch-screen devices, sketching has become more accessible and therefore has received increasing attention. Deep learning has emerged as a potential tool for SBIR, allowing models to automatically extract image features and learn from large amounts of data. To the best of our knowledge, there is currently no systematic literature review (SLR) of SBIR with deep learning. Therefore, the aim of this review is to incorporate related works into a systematic study, highlighting the main contributions of individual researchers over the years, with a focus on past, present and future trends. To achieve the purpose of this study, 90 studies from 2016 to June 2023 in 4 databases were collected and analyzed using the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) framework. The specific models, datasets, evaluation metrics, and applications of deep learning in SBIR are discussed in detail. This study found that Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN) are the most widely used deep learning methods for SBIR. A commonly used dataset is Sketchy, especially in the latest Zero-shot sketch-based image retrieval (ZS-SBIR) task. The results show that Mean Average Precision (mAP) is the most commonly used metric for quantitative evaluation of SBIR. Finally, we provide some future directions and guidance for researchers based on the results of this review.

Crime remains a crucial concern regarding ensuring a safe and secure environment for the public. Numerous efforts have been made to predict crime, emphasizing the importance of employing deep learning approaches for precise predictions. However, sufficient crime data and resources for training state-of-the-art deep learning-based crime prediction systems pose a challenge. To address this issue, this study adopts the transfer learning paradigm. Moreover, this study fine-tunes state-of-the-art statistical and deep learning methods, including Simple Moving Averages (SMA), Weighted Moving Averages (WMA), Exponential Moving Averages (EMA), Long Short Term Memory (LSTM), Bi-directional Long Short Term Memory (BiLSTMs), and Convolutional Neural Networks and Long Short Term Memory (CNN-LSTM) for crime prediction. Primarily, this study proposed a BiLSTM based transfer learning architecture due to its high accuracy in predicting weekly and monthly crime trends. The transfer learning paradigm leverages the fine-tuned BiLSTM model to transfer crime knowledge from one neighbourhood to another. The proposed method is evaluated on Chicago, New York, and Lahore crime datasets. Experimental results demonstrate the superiority of transfer learning with BiLSTM, achieving low error values and reduced execution time. These prediction results can significantly enhance the efficiency of law enforcement agencies in controlling and preventing crime.

Smart city infrastructure has a significant impact on improving the quality of humans life. However, a substantial increase in the urban population from the last few years poses challenges related to resource management, safety, and security. To ensure the safety and security in the smart city environment, this paper presents a novel approach by empowering the authorities to better visualize the threats, by identifying and predicting the highly-reported crime zones in the smart city. To this end, it first investigates the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) to detect the hot-spots that have a higher risk of crime occurrence. Second, for crime prediction, Seasonal Auto-Regressive Integrated Moving Average (SARIMA) is exploited in each dense crime region to predict the number of crime incidents in the future with spatial and temporal information. The proposed HDBSCAN and SARIMA based crime prediction model is evaluated on ten years of crime data (2008-2017) for New York City (NYC). The accuracy of the model is measured by considering different time scenarios such as the year-wise, (i.e., for each year), and for the total considered duration of ten years using an 80:20 ratio. The 80% of data was used for training and 20% for testing. The proposed approach outperforms with an average Mean Absolute Error (MAE) of 11.47 as compared to the highest scoring DBSCAN based method with MAE 27.03.

The continued urbanization poses several challenges for law enforcement agencies to ensure a safe and secure environment. Countries are spending a substantial amount of their budgets to control and prevent crime. However, limited efforts have been made in the crime prediction area due to the deficiency of spatiotemporal crime data. Several machine learning, deep learning, and time series analysis techniques are exploited, but accuracy issues prevail. Thus, this study proposed a Bidirectional Long Short Term Memory (Bi-LSTM) and Exponential Smoothing (ES) hybrid for crime forecasting. The proposed technique is evaluated using New York City crime data from 2010-2017. The proposed approach outperformed as compared to state-of-the-art Seasonal Autoregressive Integrated Moving Averages (SARIMA) with low Mean Absolute Percentage Error (MAPE) (0.3738, 0.3891, 0.3433,0.3964), Root Mean Square Error (RMSE)(13.146, 13.669, 13.104, 13.77), and Mean Absolute Error (MAE) (9.837, 10.896, 10.598, 10.721). Therefore, the proposed technique can help law enforcement agencies to prevent and control crime by forecasting crime patterns.

Many recent studies have shown that various multi-objective evolutionary algorithms have been widely applied in the field of search-based software engineering (SBSE) for optimal solutions. Most of them either focused on solving newly re-formulated problems or on proposing new approaches, while a number of studies performed reviews and comparative studies on the performance of proposed algorithms. To evaluate such performance, it is necessary to consider a number of performance metrics that play important roles during the evaluation and comparison of investigated algorithms based on their best-simulated results. While there are hundreds of performance metrics in the literature that can quantify in performing such tasks, there is a lack of systematic review conducted to provide evidence of using these performance metrics, particularly in the software engineering problem domain. In this paper, we aimed to review and quantify the type of performance metrics, number of objectives, and applied areas in software engineering that reported in primary studies—this will eventually lead to inspiring the SBSE community to further explore such approaches in depth. To perform this task, a formal systematic review protocol was applied for planning, searching, and extracting the desired elements from the studies. After considering all the relevant inclusion and exclusion criteria for the searching process, 105 relevant articles were identified from the targeted online databases as scientific evidence to answer the eight research questions. The preliminary results show that remarkable studies were reported without considering performance metrics for the purpose of algorithm evaluation. Based on the 27 performance metrics that were identified, hypervolume, inverted generational distance, generational distance, and hypercube-based diversity metrics appear to be widely adopted in most of the studies in software requirements engineering, software design, software project management, software testing, and software verification. Additionally, there are increasing interest in the community in re-formulating many objective problems with more than three objectives, yet, currently are dominated in re-formulating two to three objectives.

Human Activity Recognition (HAR) is an active research area due to its applications in pervasive computing, human-computer interaction, artificial intelligence, health care, and social sciences. Moreover, dynamic environments and anthropometric differences between individuals make it harder to recognize actions. This study focused on human activity in video sequences acquired with an RGB camera because of its vast range of real-world applications. It uses two-stream ConvNet to extract spatial and temporal information and proposes a fine-tuned deep neural network. Moreover, the transfer learning paradigm is adopted to extract varied and fixed frames while reusing object identification information. Six state-of-the-art pre-trained models are exploited to find the best model for spatial feature extraction. For temporal sequence, this study uses dense optical flow following the two-stream ConvNet and Bidirectional Long Short Term Memory (BiLSTM) to capture long term dependencies. Two state-of-the-art datasets, UCF101 andHMDB51, are used for evaluation purposes. In addition, seven state-of-the-art optimizers are used to fine-tune the proposed network parameters. Furthermore, this study utilizes an ensemble mechanism to aggregate spatial-temporal features using a four-stream Convolutional Neural Network (CNN), where two streams use RGB data. In contrast, the other uses optical flow images. Finally, the proposed ensemble approach using max hard voting outperforms state-of-the-art methods with 96.30% and 90.07% accuracies on the UCF101 and HMDB51 datasets. [ABSTRACT FROM AUTHOR]