For CNN image recognition tasks, like object recognition/face recognition/object segmentation/posture recognition, are there experiment results about how much will the performance be degraded with monochrome images?

The imaginary experiment is like:

  1. Take the existing frameworks, reduce the channel number in the framework to fit monochrome images

  2. Transform the existing training data and testing data to monochrome images

  3. Train the model with monochrome training data.

  4. Use the model to test the monochrome testing data.

  5. Compare the result with the original result.

  • $\begingroup$ By "Compare the result with the original result", you assume that another CNN has been trained with the original images, so that we can compare it with the CNN trained with monochrome images, right? $\endgroup$ – nbro Jun 24 at 11:58
  • $\begingroup$ @nbro Yes, everything begin with an existing successful CNN model for color image. $\endgroup$ – jw_ Jun 24 at 14:36

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.