We are using ML.NET to classify images. Every image should fall into one of more classes. The images contain different variants of products. The variants of the products have lots of similarities but also some distinct features.
Is it possible to add the variant of the product as a feature to the learning algorithm which then can be passed along with to the prediction function? So training can make use of the similarities between the different product variants but we can still give a hint to the prediction as we know beforehand which product variant is being predicted?
Or do i have to train a separate model for every variant of the product? I am doing this right now as i haven't figured out how to pass additional features besides the image data.
The ImageClassification task documentation suggests that The feature column must be a variable-sized vector of Byte..
The documentation on how to improve my model suggests to Add context to the data.
Is it possible to add additional feature columns along with the image data or is this not possible with ML.NET (or in general?).