Machine learning has sneaked into our lives. Every night I tell Alexa "turn the lights off". Every day my phone successfully performs face recognition before I pay with it. Given the tremendous improvement of this field I decided to find out how hard nowadays it is for an ordinary frontend engineer to apply machine learning for some task.
A long, long ago my friend and me used to share photos of what we had eaten during the day. We tried to eat less and make healthier food choices. So I decided to try to make something (a website? some bot? to be refined later), which helps the user to tweet about his recent food choices. Show it the food photos and it creates a tweet code with suitable emojis which you can share. Thanks to Wordle for the inspiration :)
Of course, being a former researcher the first thing I did was to google what other people have done about food recognition. The most interesting article I found so far is here
I decided to kind of repeat the author's way, but give it a bit of my own tweak. E.g. most probably I'll use 3 labels: pastry, salad and soup.
Next post is going to be about how I started to gather the dataset for my project.
Top comments (0)