Would like to know whether a proposed method can have different accuracy percentages when it is tested on two different data sets like HRF and DRIONS-DB? What are the factors that determine it if so?
The accuracy varies in a general way in image classification when tested on different datasets which depend on different reason like the quality of images, preprocessing methods which you are used for enhancement the images and also it depends on which the type of feature you extract. all of these can be the effect on the accuracy of any system.
My suggestions to you before starting to test, check the quality measure of all the datasets and see which enhancement technique suitable for each datasets these will be done after some experimental work and make the comparative study of that. Afterward start your next stages. and you will be able to see why the results of accuracy was varied from each other