📖💡🤔 Top 6 Datasets for Computer Vision
CIFAR-10 - contains 60 thousand 32x32 color images with 10 classes (animals and real world objects). Each class consists of 6000 images. This dataset contains 50,000 training images and 10,000 test images. The classes are mutually exclusive and there is no overlap between them.
Kinetics-700 - This is a large array of videos, consisting of 650,000 clips describing 700 classes of human actions. The video includes human-thing interactions such as playing musical instruments and human-human interactions such as hugging. Each activity class contains at least 700 video clips, and each clip is annotated with an activity class that is longer than 10 seconds.
IMDB-Wiki - One of the largest publicly available datasets of human faces with gender, age and name. In total it contains 523051 images, 460723 faces are photos of 20284 celebrities from IMDb and 62328 celebrities from Wikipedia.
Cityscapes is a database containing a diverse set of stereographic video clips recorded on the streets of fifty cities. The clips were filmed for a long time under various lighting and weather conditions. Cityscapes contains semantic segmentation of object instances with pixel precision for 30 classes divided into 8 categories. It provides 5000 frame per pixel annotation and 20000 frame approximate annotation.
Fire and Smoke Dataset - This is a dataset of over seven thousand unique HD resolution images. It consists of photos of starting fires and smoke taken by mobile phones in real situations. The pictures were taken in a wide range of lighting and weather conditions. This data set can be used for fire and smoke recognition and detection, as well as anomaly recognition.
FloodNet Dataset - This array consists of high resolution images taken from unmanned drones. The images contain detailed semantic annotations of damage caused by hurricanes.