We are going to use medicinal plant leaf image dataset (https://www.kaggle.com/datasets/bijlyborkatullah/medicinal-plant-bijly-withbg/data) from Kaggle to illustrate the image analysis and identification using TensorFlow(TF)/Keras here.
We are going to develop two(2) different kind of models using TF/Keras for the medicinal plant leaf recognition. These models commonly use keras image pre-processing utility for image dataset reshaping and pre-processing.
The first(1st) model is a simple CNN/DNN model and offers flexibility for you to play around optimizing the layers and weights to improve the training and validation accuracy while minimizing their losses respectively.
The second(2nd) model is based on Transfer Learning where we are going to use InceptionV3 pre-trained model as the base and build custom DNN layers on top of it to quickly arrive at an optimized training/validation metrics.
Both of these models can be downloaded and optimized further with additional data (or) you can try using it with a new image dataset for a different use case as well.