100 classes,total 800 images, which deep learning model do i need

I have 100 classes of 800 images and i want to train a classification model. Which deep learning model is suitable? need your advice~

800 images for each class or 800 in total? In the latter case you only have 8 images per class on average which probably isn’t enough.

The data doesn’t seem enough but what are the images? It may have a chance to work if the images are similar to another larger open source corpus available so that you can train on that before training on your images. Similar to how pretrained imagenet weights work really well on cats and dogs since the images are similar.

If such a dataset is available and you have decent compute, you could either use leave one out cross validation to select the best model architecture or pick the one that performs the best on the bigger dataset