1
$\begingroup$

I have around 1600 images extracted from videos shot at night time. I am labeling each image and trying to be as accurate as I can in assigning bounding boxes. I am labeling vehicles and traffic light/traffic signs. This is very time-consuming, I am wondering if someone have experience or done this before and can advise me on some automated methods of labeling night time images. The objects of interests are usually appear to be quite small. An example labelled image is given below

enter image description here

$\endgroup$
1
  • $\begingroup$ use amazon mechanical turk perhaps $\endgroup$ – Van Peer May 26 '20 at 5:58
1
$\begingroup$

On the internet, LabelImg is pretty popular. It's a Python program that you can use to automatize drawing of bounding boxes. Alternatively there is HyperLabel, on their website they say it's free so it.s worth to give it a try.

The options available are so many I completely get lost while searching. I suggest you to read this nice review of tools first.

With these tools, labeling 1600 images is not going to be a problem IMHO. Good luck.

$\endgroup$
1
  • $\begingroup$ I find all these tools are for manually labeling images, I have used LabelImg. You are right 1600 images are not that much of work, but I am getting more and more image data. Therefore, I do not want to stick to manual labeling, so specifically my question is how can I automate image labeling process? $\endgroup$ – Vendetta May 26 '20 at 7:02
1
$\begingroup$

If you're drawing bounding boxes around cars, you can use a pretrained model. Hundreds of pretrained networks are trained on the COCO dataset, which has a car category. Then you will get the labels as the output.

Then you can also double check with LabelImg by iterating through the folder with the keyboard arrows, and adjust the bounding boxes if they are off. But they shouldn't. Cars (and people) are maybe the most cropped category in the world. Use it to your advantage.

$\endgroup$
2
  • $\begingroup$ Could you please give more information on "you can use a pretrained model"? How exactly I can use a pretrained model to obtain labels as the outputs. Can you please provide a link to some example where it has been done? $\endgroup$ – Vendetta May 26 '20 at 11:45
  • $\begingroup$ check for transfer learning where you can use a pretrained model to get the labels for your own images. $\endgroup$ – ARAT Feb 23 at 19:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.