SAIOD Sikkim Aerial Images dataset for Object Detection
Description
The SAIOD dataset is an aerial image dataset prepared by collecting images from Sikkim state located in the Himalayas, India, using a DJI Mavic Air drone. The terrain is diverse, including undulating surfaces, slopes, gorges, rivers, forests, and settlements. Ten classes of objects were identified for image classification, including buildings, cars, debris, footpaths, metalled roads, open fields, shadows, tanks, trees, and roofs.
Files
Steps to reproduce
Over 300 high-resolution images were captured from heights ranging between 60 to 120 meters above the ground, primarily during daylight but also under low light conditions. Ten classes of objects were identified, primarily for image classification task, that includes: buildings, cars, debris, footpaths, metalled roads, open fields, shadows, tanks, trees, and roofs. The dataset comprises of sliced images of 270x270 resolution with a 50% overlap generated from the acquired aerial images from drone. Manual selection of images was performed to ensure sufficient ground truth information for each class. A total of 12,500 image slices (1250 per class) are selected. The dataset has been divided using a random indexing function into 80% training images (1000 images) and 20% testing images (250 images) sets for each class.