Image dataset with timestamps?

Question

I am starting an exploratory project to determine the time (of day) of which a photo is taken based on image features. I've done some research and downloaded a few databases hoping that "creation date", "last modified", or the file names will properly reflect the time when the original photo was taken, but I have had no luck.

So far I will be resorting to all of my personal photos, which I do not think will be sufficient enough in quantity or diverse enough in quality.

Are there any natural image datasets out there with timestamps or even timestampped file names?

Answer 1

I found a dataset with EXIF data! It's the MIRFLICKR dataset. I was looking for more amateur-style images, but this is a great start.

I'll post if I make any more findings. Other datasets that I've checked have some images with EXIF data, but they've ranged from 1 in 100 to less than 1 in 1000 images with useful EXIF data.

Answer 2

I am sure you would have access to video files. Pick any of them and split it in numerous frames. Simultaneously capture the timestamp as well. This is just one way of creating your own image dataset with timestamps.

If you've access to MATLAB, this shall help you:

vidObj = VideoReader('xylophone.mp4');
timeStamps = [];
while hasFrame(vidObj)
    timeStamps(end+1) = vidObj.CurrentTime;
    img = readFrame(vidObj);
    imwrite(img,        strcat('xylophone_',num2str(numel(timeStamps)),'.png'));
end

If Python comforts you, (using OpenCV) then you may try:

vidcap = cv2.VideoCapture('some_video_file.mp4')
success,image = vidcap.read()
count = 0
success = True
while success:
  cv2.imwrite("frame%d.jpg" % count, image)
  success,image = vidcap.read()
  count += 1

Timestamp is something you need to figure out, for instance if you wish to capture system time of when the frame is getting extracted from video file, or may opt for something else.