The user to be able to take a picture of their plant leaf and have detecting any possible disease from that.
- FastAI - For Data Reading & Preprocessing & Training DL Model
- FastAPI - For wrapping the DL Model into an API an send it to Backend Team
The Dataset has over 38 Different Classes and 87k RGB image. And Below if my implementation -
- Used Color Images ( 3 Channel )
- No Data Argumentation
- Data Notmalisation
- Batch Size = 100
- Training 90%, Validation 10%
The model is Resnet 18 (Paper).
With Training the model for 1 epoch, took 16 minutes in single Tesla P100 GPU (I use Kaggle Kernel ), with accuracy as metrics
- Training Loss = 0.118667
- Validation Loss = 0.066453
- Accuracy = 0.977258
Distributed under the Apache License 2.0. See LICENSE for more information.