SD-3000
Description
The purpose of the dataset selection was to train a YOLO-based machine learning model for shadow detection. In surveillance systems, where accurate and timely shadow identification is crucial to enhancing security standards, this approach finds extensive application. The focus of the collection is on several shadow classes, including human, dog, cow, and gun shadows, which makes it flexible enough to be used in a range of surveillance scenarios. • Human Shadow: The SD-3000 Dataset consists of 956 images associated with human shadows. Among these, 669 images are allocated for training, 143 for testing, and another 143 for validation. • Dog Shadow: For the dog shadow class, there are 877 images in total. Out of these, 614 are designated for training, 132 for testing, and an additional 132 for validation. • Cow Shadow: The cow shadow class is represented by 357 images in the dataset. Among these, 250 images are earmarked for training, 54 for testing, and an identical 54 for validation. • Gun Shadow: The gun shadow class contributes 810 images to the dataset. Of these, 567 are utilized for training, 122 for testing, and an equivalent 122 for validation.