So I've created a model to classify images of traffic signs. I'm currently using this model to predict classes of images I've gotten from the internet. However, I would like to be able to re-classify or correctly label images in which the model failed to correctly predict their classes. One way would be to place the images in their correct classes and retrain the model from the last saved checkpoint. I don't think this is a good approach as the model should overfit over time. Hence, I'd like to know if there's a better way to achieve this.
There are two basic approaches to this problem:
The performance of the old model was not good enough, you are able to gather much more data and train a new better model
You setup an "online model" which continuously learns and improves but is setup very differently.
Do you have a specific performance level in mind that you need / want to hit? In that case simply gathering more data and retraining may be fine. However if you simply want to continuously improve you need to approach it with an "online model".