When I am reading papers, I am finding researchers explain some methods in machine learning.
They judged this or that method is suitable for high-dimensional images.
I would like to know how I can distinguish between high and low-dimensional images.
Is there any threshold or else to know high and low-dimensional images?
In deep learning, training on a data set of images can be very time consuming, therefore, many published works rely on training on small patches (say 64x64). As such, images with 512x512 can be a heavy burden to the system when training.
There is no agreed upon cut for what is high/low dimensional image.
It depends on machine power, HPC and Google's GPU-enabled servers for example have no issue dealing with a 512x512 or even higher dimensional images to train their deep learning models.
N.B: I understood your question as addressing image sizes (dimensionality).
Thank you so much for your answers Abbas Cheddad and Aparna Sathya Murthy
- Badges
- Science topic