I am training a deep learning model for object detection. The consensus is that the more images that you have, the better the results will be. All the tutorials that I have seen say that more images are key.

I am labeling objects in my images with Label-Img, which provides the algorithm with specific training samples on the images. For my images, I am using photos with dimensions of 1100 x 1100 pixels. In my case, I could generate anywhere between 50-100 high-quality training samples per image. For example:

enter image description here

In cases such as this where large numbers of training samples can be generated from a single image, do you really need several hundred images? Or can you lessen the number of images because of the number of training samples?

  • $\begingroup$ The short answer could be "yes, if a single image contains several instances of the object, that made thinks easier and number of training images can be reduced". However, for a good answer, could you describe what kind of NN network you use ? $\endgroup$ – pasaba por aqui Nov 7 '20 at 10:27
  • $\begingroup$ @pasabaporaqui Convolutional Neural Networks (i.e. ResNet50) and RCNNs. $\endgroup$ – ihb Nov 8 '20 at 21:20

Your Answer

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

Browse other questions tagged or ask your own question.