SURGICAL TOOLS
Description
This dataset was created to advance computer vision and object detection systems for surgical tool recognition in clinical settings. The research hypothesis focuses on enhancing the accuracy and reliability of machine learning models to identify surgical tools under diverse, real-world conditions. By compiling a robust image collection that reflects the complexities of surgical environments, the dataset aims to minimize overfitting and boost model performance for practical applications. It consists of 6,000 high-quality images, with 5,000 manually captured at Amrita Vishwa Vidyapeetham, Chennai, India, during January and February 2025, and 1,000 sourced from various online platforms to add variability. The dataset spans nine categories of surgical tools: forceps, hemostats, scalpels, mayo scissors, syringes, bandage scissors, episiotomy scissors, surgical gloves, and medical cotton, offering a broad representation of tools used in surgeries. Notable findings include a significant subset of 1,520 images showcasing both overlapping and non-overlapping tool configurations, mimicking real surgical scenarios. This is vital for training models to manage cluttered or complex arrangements. The dataset also captures tools under diverse conditions—blur, artificial blood, varying backgrounds, lighting from dim to bright, and 360-degree angles—ensuring exposure to challenges encountered in operating rooms.
Files
Steps to reproduce
The dataset was assembled through a detailed process to ensure quality and diversity, allowing replication by others. The core portion—5,000 images—was manually captured at Amrita Vishwa Vidyapeetham, Chennai, India, in January and February 2025. A Vivo T2 Pro smartphone, chosen for its high-resolution camera, was used to photograph surgical tools in controlled settings. Tools were arranged in various setups, including overlapping and non-overlapping configurations, and captured from 360-degree angles for comprehensive coverage. To reflect real-world conditions, variables like intentional blur (simulating motion), artificial blood on tools, and lighting adjustments (dim to full brightness) were introduced. Backgrounds varied using surgical drapes and trays to mimic operating room diversity. Surgical professionals supervised the process, ensuring accurate tool positioning and clinical relevance. The cameras core is a 64-megapixel primary sensor, featuring an f/1.8 aperture and a 26mm wide-angle lens, which strikes a balance between light intake and a broad field of view, essential for detailed photography of intricate surgical instruments. This sensor, likely a Sony IMX682 or a comparable mid-range option (though not explicitly specified by Vivo), boasts an approximate 1/1.72-inch sensor size and 0.8 μm pixel size, enabling it to capture fine details even under varying conditions.