2
$\begingroup$

I am experimenting with Facial Expression Recognition and want to use a pretrained CNN model and a multi-stage fine tuning strategy to deal with scarce data. I came across the work of Knyazev et al. (2017). The authors fine tune a VGG-Face Face Recognition model on the FER2013 Facial Expression Recognition dataset. Unfortunately, they don't explain how they fine-tune a model trained on RGB images on a grayscale image dataset.

Is it possible (and practical) to fine tune a CNN originally trained on RGB images on a target dataset consisting of grayscale images? What is the general approach?

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ I found a similar question on Stackoverflow - How can I use a pre-trained neural network with grayscale images?. The suggestion is to convert the grayscale images to RGB. But I imagine this will greatly alter the input distribution and can't imagine it working well. Could anybody recommend literature on the problem? $\endgroup$
    – Boyan Hristov
    Commented Apr 20, 2021 at 17:36

1 Answer 1

Reset to default
0
$\begingroup$

I think the solution is to resize the images into a format that is no smaller than 197*197, and to recolor the images also.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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