Search

Automatic electronic waste management needs to classify components specifically from laptops of different models and brands so that parts can be recycled efficiently using AI-enabled robots. To achieve this goal, a good dataset plays a significant role as the systems that operate e waste management machines will learn from this dataset and act accordingly. The lack of proper datasets that are available publicly related to components of any type of device can be a barrier to the work process. This research aims to improve the scenario by developing, to the best of knowledge, the first-ever publicly available and ready-to-use standard dataset of laptop components. Most of the images are collected from three different computer firms. The dataset consists of 29,120 images of 26 different components. The dataset is developed using components of laptops from eight models of HP brand and it deals with the common components that can be found in any laptops of any brand. This dataset is anticipated to have a significant impact on the field of automated recycling as it is expected to attract wide attention from machine-learning researchers and practitioners in this specified field.

This article introduces a comprehensive dataset developed in collaboration with the Bangladesh Institute of Nuclear Agriculture (BINA) and the Bangladesh Rice Research Institute (BRRI), featuring high-resolution images of 38 local rice varieties. Captured using advanced microscopic cameras, the dataset comprises 19,000 original images, enhanced through data augmentation techniques to include an additional 57,000 images, totaling 76,000 images. These techniques, which include transformations such as scaling, rotation, and lighting adjustments, enrich the dataset by simulating various environmental conditions, providing a broader perspective on each variety. The diverse array of rice strains such as BD33, BD30, BD39, among others, are meticulously detailed through their unique characteristics—color, size, and utility in agriculture—providing a rich resource for research. This augmented dataset not only enhances the understanding of rice diversity but also supports the development of innovative agricultural practices and breeding programs, offering a critical tool for researchers aiming to analyze and leverage rice genetic diversity effectively.

In Bangladesh, there are significant number of medicinal plants, but currently no comprehensive record of these valuable species is publicly available. Alarmingly, some of these plants are in a precarious state of endangerment. Therefore, we are creating a unique dataset of Bangladeshʼs rare, endangered, and threatened medicinal plants to support conservation efforts. It will help us to track and conserve endangered plant species, ensuring a more organized approach to research and preservation efforts. We conducted on-site visits to the National Botanical Garden and The Government Unani and Ayurvedic Medical College, capturing photographs of these plants in optimal sunlight conditions at various times of the day. This involved fieldwork, detailed image annotations, dataset organization, diversity augmentation, and contribution to the preservation of our natural heritage. We have collected a total of 16 types of rare and endangered medicinal plant leaf photos to create our unique dataset consisting of a total of 3494 images. This dataset will help researchers in biodiversity conservation through building efficient machine learning models and applying advanced machine learning techniques to identify rare and endangered medicinal plants.

The agricultural sector is a vital component of Bangladesh's economy, but its agri-food supply chain faces signifi-cant inefficiencies primarily due to the involvement of numerous intermediaries. This complexity not only reduces the profits for farmers but also affects the overall transparency and efficiency of the supply chain. This study aims to em-ploy blockchain technology to transform the traditional agri-food supply chain in Bangladesh, focusing on increasing transparency, enhancing efficiency, and improving profitability for farmers, thus potentially bolstering the entire agri-food ecosystem in the country. The research involves setting up a blockchain-based smart contract on the Ethereum Blockchain network. This approach guarantees that all transactions within the agri-food supply chain are transparent, traceable, and accountable. Additionally, the study develops a web application to facilitate user interaction with the smart contract, enhancing accessibility and usability. Performance analysis and testing of the implemented smart con-tract demonstrate its capability to handle a significant volume of transactions without compromising on performance. The solution effectively reduces the dependency on intermediaries, thereby increasing the profit margins for the farm-ers involved. The integration of blockchain technology in the agri-food supply chain has shown promising results in enhancing transparency and efficiency. It lays a solid foundation for future improvements and suggests a scalable model that could be applied to other sectors within the country to further enhance agricultural practices and economic growth.

In the field of agriculture, particularly within the context of machine learning applications, quality datasets are essential for advancing research and development. To address the challenges of identifying different mango leaf types and recognizing the diverse and unique characteristics of mango varieties in Bangladesh, a comprehensive and publicly accessible dataset titled “BDMANGO” has been created. This dataset includes images essential for research, featuring six mango varieties: Amrapali, Banana, Chaunsa, Fazli, Haribhanga, and Himsagar, which were collected from different locations. The images were captured using the rear cameras of a Google Pixel 6a and an iPhone XR and were stored in 640 × 480 pixels resolution. Both sides of each mango leaf were photographed against white background to accurately reflect real-world scenarios in mango cultivation fields. The white background was specifically chosen to remove noise in image sample, allowing for accurate feature extraction by machine learning algorithms. This will ensure the trained model's efficacy in identifying a specific mango leaf while implemented alongside any segmentation algorithm. Additionally, image augmentation techniques such as rotation, horizontal flip, vertical flip, width shift, height shift, shear range, and zooming were applied to expand the dataset from 837 original images to a total of 6696 images (837 original image and 5859 augmented images). This expansion significantly enhances the dataset's utility for training, testing, and validating machine learning models designed for classifying mango leaf varieties, thereby supporting research efforts in this domain.

Bananas are among the most widely consumed fruits globally due to their appealing flavor, high nutritional value, and ease of digestion. In Bangladesh, bananas hold significant agricultural importance, being one of the most extensively cultivated fruits in terms of land coverage and ranking third in production volume. The banana image dataset presented in this article includes clear and detailed images of four common banana varieties in Bangladesh: Sagor Kola (Musa acuminate), Shabri Kola (Musa sapientum), Bangla Kola (Musa sp.), and Champa Kola (Musa sapientum), as well as four key stages of banana ripeness: Green, Semi-ripe, Ripe, and Overripe. The bananas were collected from wholesale markets and retail fruit shops located in different places in Bangladesh. Overall, the dataset has 2471 original images of different varieties of bananas and 820 original images of varying ripeness stages of bananas. All the images were carefully captured using a high-quality smartphone camera. Later, each image was manually reviewed, maintaining the quality standard throughout the dataset. The augmented version of the banana variety classification dataset contains 7413 images and the augmented banana ripeness stages dataset contains 2457 images. The dataset possesses immense potential in driving innovation and development of automated and efficient processes and mechanisms in several fields, including precision agriculture, food processing, and supply chain management. Machine Learning (ML) and Deep Learning (DL) models can be trained on this dataset to accurately categorize banana varieties and determine their ripeness stages. Such ML and DL models can be leveraged to develop automated systems to determine the optimal harvest time, establish standards for quality control of bananas, develop products and marketing strategies through analysis of consumer preferences for various banana varieties and ripeness levels, and streamline the banana supply chain through improvements in harvesting, sorting, packaging, and inventory management. Additionally, researchers aiming to contribute to developing Computer Vision technologies in food and agricultural sciences will find this dataset valuable in advancing precision farming and food processing mechanisms. Therefore, the dataset has a vast capacity for automating banana production and processing, minimizing the costs of manual labor, and improving overall efficiency.

The utilization of computer vision techniques has significantly enhanced the automation processes across various industries, including textile manufacturing, agriculture, and information technology. Specifically, in the domain of textile manufacturing, these techniques have revolutionized the detection of fiber defects and the quantification of cotton content in fabrics. Traditionally, the assessment of cotton percentages was a labor-intensive and time-consuming process that relied heavily on manual testing methods. However, the adoption of computer vision approaches requires a comprehensive dataset of fabric samples, each with a known cotton percentage, to serve as training data for machine learning models. This paper introduces a novel dataset comprising 1300 original images, covering a wide range of cotton percentages across thirteen distinct categories, from 30% to 99%. By employing image augmentation techniques, such as- rotation, horizontal flip, vertical flip, width shift, height shift, shear range, and zooming, this dataset has been expanded to include a total of 27,300 images, thereby enhancing its utility for training and validating computer vision models aimed at accurately determining cotton content in fabrics. Through the extraction of pertinent features from the images of fabrics, this dataset holds the potential to significantly improve the accuracy and efficiency of computer vision-based cotton percentage detection.

Criminal Activities and Crime Monitoring System has long been a research topic. In this paper, we propose an effective Crime Monitoring System (CMS) that can detect a crime in real-time using a camera surveillance system and notify the appropriate law enforcement officer. The CMS was proposed to counterbalance human weaknesses such as inattention, slow reaction, and slacking, for example, in detecting crimes. The proposed CMS detects crime scenes by combining the mechanisms and functionalities of closed-circuit television (CCTV) cameras with various deep-learning methods and image-processing techniques. The CMS operates in three stages: weapon detection, violence detection, and face recognition. We used transfer learning models to detect weapons, and violence and used a face recognition algorithm to recognize faces. More specifically, to detect weapons YOLOv5, and MobileNetv2 to detect violence and used face recognition algorithms to recognize faces. The image dataset is used in CMS, while the video dataset is used to train the MobileNet-based violence detection model. In this case, frame-by-frame images extracted from video files were used to train the model. All of the models performed admirably. The weapon detection model detected four different weapon classes with greater than 80% accuracy. The violence detection model is also 95% accurate. The face recognition model had a 97% accuracy rate in detecting faces. The CMS's combined model was tested in a variety of real-world scenarios, and its performance was found to be outstanding. It was able to detect crime incidents and generate timely alarms, demonstrating its effectiveness in providing security and safety.

Fruit production plays a significant role in meeting nutritional needs and contributing to the lessening of the global food crisis. Plant diseases are quite a common phenomenon that hampers gross production and causes huge losses for farmers in tropical South Asian weather conditions. In context, early-stage detection of plant disease is essential for healthy production. This research develops LeafNet, a convolutional neural network (CNN)-based approach to detect seven of the most common diseases of mango using images of the leaves. This model is trained specially for the pattern of mango diseases in Bangladesh using a novel dataset of region-specific images and is classified for almost all highly available mango diseases. The performance of LeafNet is evaluated with an average accuracy, precision, recall, F-score, and specificity of 98.55%, 99.508%, 99.45%, 99.47%, and 99.878%, respectively, in a 5-fold cross-validation that is higher than the state-of-the-art models like AlexNet and VGG16. LeafNet can be helpful in the detection of early symptoms of diseases, ultimately leading to a higher production of mangoes and contributing to the national economy.