So I have a dataset of 6 different dermatology disease pictures along with data of the age, gender, etc other data for each picture.
I want to train a model that combines the image and the categorical/text data to classify it as one of the 6.
My initial thought is to use ResNet to get a 255x1 size vector for each image, then make 255 features with the value in each entry is the index in the vector. Then I can just add the 5 (let's say) features to get a final input size of 300x1 for each datapoint, which I can then train an LGBM classifier or any other classifier on. Does this sound like it would work?