If I understand you question correctly, you are working on a machine learning project and part of the objective is to detect objects of silver materials, say silver ball, silver spoon etc.
However, before you can do that, you have to train your machine to differentiate objects of different materials, say silver, glass, wood etc. This is called supervised training or supervised learning in AI.
Generally, the approach towards building a supervised training AI project is as follows:
1. Collect a lot of data with known categories, in your case, images of objects of known but different materials the quantity of each should be about the same so as to maintain data balance;
2. Split the data into a training set and test set, say 60% training set and 40% test set;
3. Train the machine learning model on the training set;
4. Test the trained model on the test set to measure the accuracy.
There are many supervised training AI techniques that you can use to achieve this, ANN and SVM are just some of them.
Since you are interested in OpenCV, there are no better place than to visit its official
website for more information.