Hey everyone! I'm diving into a project that involves a classifier model for computer vision. My goal is to use an image as input and get a list of the ingredients present in that meal. Currently, I'm trying out one of Clarifai's models, but I'm not thrilled with its accuracy—like, for a picture of chicken breast, it only recognizes it as 'meat' or 'chicken.' I'm looking for suggestions on better models or libraries, and whether I should go with open-source options or pay-per-call APIs. Any thoughts?
3 Answers
Check out this GitHub project called NotHotdog-Classifier! It might give you some insights or a good starting point for your model. You can find it at github.com/kmather73/NotHotdog-Classifier.
Is this related to Python in some way? Just curious if the ML side of things ties into that.
Honestly, this might be a tough nut to crack. A model or even a human can't really tell what's in a dish just by looking at it, especially seasonings and spices. Plus, many dishes look similar, which makes it even harder for any model to learn the specifics of what ingredients correspond to what dish.
I agree, it's basically impossible. So much about a meal comes from smell and taste! You can't tell if lemon juice or salt is in a dish just by seeing it unless there's way too much. And think about the ingredients that are cooked off or discarded—visual cues just won't cut it.
Not the OP, but I’d say Python is definitely a go-to language for machine learning!