Full Stack Deep Learning: Project Deep Shroom
A North EU mushroom image classifier trained on a kaggle dataset with fastai
Full Stack Deep Learning: Project Deep Shroom | May, 2023 | Kaggle Notebook (train & save model)
Using fast.ai, trained a ResNet18 model on a Kaggle dataset consists of 9 different folders that contains from 300 to 1500 selected images of corresponding mushrooms genuses, labels indicated by folder names. Deployed and published using Gradio on Hugging Face: click here. Achieved accuracy of 0.8323 and near perfect top 3 accuracy of 0.9627.