DATS_2020

The dataset contains more than 9,000 images. A mix of photo and video mode of the camera was used to capture the images. 97 video sequences, 10-15 sec long with 30 frames per second (FPS) were recorded on various locations at different times of the day. 7,554 images could be recovered from the sequences as there were lot of redundant images. VLC Media Player was used to extract the frames from the video sequences. The images have a resolution of 1088 x 1920. Images collected in photo mode are 1,524 having a resolution of 2134 x 2988. 47 different classes are identified currently in this set of images.本数据集共计包含逾9000张图像，采集过程混用了相机的拍照与录像模式。共录制97段视频序列，每段时长10至15秒，帧率为30帧每秒（FPS），拍摄场景覆盖不同时段的各类地点。由于视频序列中存在大量冗余图像，最终从中提取出7554张图像，图像帧提取工作通过VLC媒体播放器完成。视频提取所得图像的分辨率为1088×1920；拍照模式采集的图像共计1524张，分辨率为2134×2988。当前该数据集已标注有47个不同类别。