I have a few general questions.
The training dataset consists of images. So against what do I have to train the model from training data? Since there is no similarity index of some sort associated with images from training images, am I supposed to alter the images myself to test or develop a variable or something like that to compare, or use the reference images, like so. In general, how is the training set to be used?
I can use any language and algorithm to solve the problem?
I have to submit my solutions for Matching track only? Not necessarily for Descriptor track too?
I look forward for a prompt and descriptive answer ASAP.
Many thanks in advance.