-1

I study deep learning using images. This time, I have a question. I want to hear everyone's opinion.

Generally, learning data is used for image recognition of the Website which reducing the data capacity. So I think jpeg images are used. However, the embedded system, the compress images aren't used. I think raw image is used. I think that raw image is saved in memory and processed for recognition.

Here, when using the jpeg images on the Website for learning model, the formats of the images are different. Is there any difference in recognition? Since raw image is larger information than the jpeg image. Will not problems occur in recognition?

Thank you for your answers.

oguririn
  • 7
  • 2

1 Answers1

0

I want to hear everyone's opinion.

Opinion based questions aren't a good fit for SO but let me try to answer your questions.

A neural network (usually) doesn't work on the compressed media data. JPGs, PNGs or GIFs are first decompressed and stored as raw 2D arrays (3D, if your work with color) before feeding them into the actual architecture.

That doesn't necessarily mean that there won't be a difference between using JPGs and RAW files. In theory, RAW files will lead to more accurate results. In practice, the difference is very minor if you don't overdo the compression. There is a paper dedicated to this but I'm having trouble finding it.

pietz
  • 2,093
  • 1
  • 21
  • 23