Hi everyone 👋

I feel like I'm a bit late to the party but wanted to submit the project nonetheless 💩

My grandparents know about everything there is to know about mushrooms and since I had been a child, I had nothing but admiration for their skills (they survived each serving after all...).

Now there are numerous apps out there for automatically recognizing mushrooms. I wanted to know whether it was possible to rebuild something similar in a reasonable amount of time and I proudly present to you the Mushroom Tester!

The app is based on a dataset that I obtained from Wildlife UK's Mushroom Guide. I used this dataset to train a model on colabel which I connected to Bubble.

I had between 5-30 images for each type of mushroom, which is not sufficient to recognize it. Given the tiny dataset (2,000 images in total), the model was only trained to recognize whether a mushroom is edible, inedible, or poisonous.