Artificial Intelligence

Annotating the scene text in the PRIVATY-TEXT-IMAGE dataset was done in Adobe Photoshop.   To maintain the rationality of the annotation operation, the images' aesthetics, and the textures' consistency around the deleted text areas, we utilized the content-aware fill feature of Photoshop.   This feature can enhance intelligent editing and modification capabilities during the image processing, automatically analyze the image content around the private text areas, and generate matching filling content to make the images look more natural and complete.  

Annotating the scene text in the PRIVATY-TEXT-IMAGE dataset was done in Adobe Photoshop.   To maintain the rationality of the annotation operation, the images' aesthetics, and the textures' consistency around the deleted text areas, we utilized the content-aware fill feature of Photoshop.   This feature can enhance intelligent editing and modification capabilities during the image processing, automatically analyze the image content around the private text areas, and generate matching filling content to make the images look more natural and complete.  

Annotating the scene text in the PRIVATY-TEXT-IMAGE dataset was done in Adobe Photoshop.   To maintain the rationality of the annotation operation, the images' aesthetics, and the textures' consistency around the deleted text areas, we utilized the content-aware fill feature of Photoshop.   This feature can enhance intelligent editing and modification capabilities during the image processing, automatically analyze the image content around the private text areas, and generate matching filling content to make the images look more natural and complete.  

The Smart Home Device Dataset consists of 5000 samples collected at an hourly interval starting from January 2022, representing consumer electronics and IoT-enabled devices in a home automation environment. Each entry is associated with a unique device ID, ensuring identification of distinct devices. The dataset captures real-time sensor readings, including temperature variations (18°C to 30°C), power consumption levels (10W to 500W), and user activity states (Active, Idle, or Sleep), which provide contextual insights into device operation.

• LIS is a low - light instance segmentation dataset. https://github.com/Linwei-Chen/LIS
• ExDark is a low - light object detection dataset. https://github.com/cs-chan/Exclusively-Dark-Image-Dataset
• ACDC is a semantic segmentation dataset for adverse weather conditions.

The Dash Cam Video Dataset is a comprehensive collection of real-world road footage captured across various Indian roads, focusing on lane conditions and traffic dynamics. Indian roads are often characterized by inconsistent lane markings, unstructured traffic flow, and frequent obstructions, making lane detection and traffic identification a challenging task for autonomous vehicle systems. Reliable lane detection is crucial for developing robust Advanced Driver Assistance Systems (ADAS) and autonomous driving models tailored for Indian conditions.

Bananas are widely farmed and consumed, offering essential nutrients like manganese, vitamin B6, vitamin C, and magnesium. They come in various breeds with distinct visual traits, including size, shape, color, texture, and skin patterns. To classify these varieties, five deep learning models—VGG16, ResNet50, MobileNet, Inception-v3, and a customized CNN—were trained on banana images. These models enhance quality control and supply chain management by accurately identifying banana breeds.

The integration of wearable sensors with artificial intelligence forms the base for analyzing physical activities through digital signal processing, numerical methods, and machine learning. Computational intelligence and communication technologies enable personalized monitoring, training, and rehabilitation, with applications in sports, neurology, and biomedicine. This paper focuses on motion analysis in alpine skiing using real accelerometric, gyroscopic, positioning, and video data to evaluate ski movement patterns.

We construct the Thyroid Nodule Ultrasound (TNUS) dataset with thyroid nodule positions and puncture annotations, lacking in existing datasets. It supports future research in automating detection and diagnosis, enhancing diagnostic accuracy and clinical applications. The TNUS dataset is a curated collection of thyroid nodule ultrasound (US) images designed to support research in puncture position detection and nodule segmentation. It contains 4,376 images with puncture position annotations and 2,626 additional images with thyroid/nodule masks.