ausafmo / cs229-collected-lecturenotes Goto Github PK

🙋‍♂️ The Training notebook(s) used to train/FineTune ResNet50 on Mendley Medicinal leaf dataset

The Code for the model serving can be found at the following repo links, There are two repos for Front-End and Back-End Components

• For Backend App see Backend Repo
• For Front-End see Training Notebook

• The Project aims to solve the problem of identification of medicinal herbs.
• The Machine Learning Model uses ResNet, with a validation accuracy of 98% and testing accuracy of 96%.
• The training and testing dataset contains over 1500 images across 50 medicinal herb species.
• The model is capable of classifying the user-given image into 30 species.

• Python 3
• Pandas
• Numpy
• Pillow
• PyTorch
• Tensorflow
• ResNet50 model trained on Mendeley Medicinal Leaf Dataset
• Dataset: S, Roopashree; J, Anitha (2020), “Medicinal Leaf Dataset”, Mendeley Data, V1, doi: 10.17632/nnytj2v3n5.1
  

• Training Accuracy : 97.55%
• Validation Accuracy : 98.64%
• Top 1 Testing Accuracy : 96.43%
• Top 5 Testing Accuracy : 99.07%