I used the Fruit 360 dataset hosted on the Kaggle website to classify fruit images into 131 different classes. I was able to get 85% accuracy on the training set of data.
I made a similar model using feedforward neural network to get an idea about how CNN outperforms feedforward neural netowk. Using the feedforward neural network the best I could get is 61% accuracy.
Note:
Before running the notebook in you computer first make sure Anaconda Navigator is installed along with other required libraries. Download the dataset on the same file where you have downloaded the ipynb file. You can also run the notebook directly on the Kaggle website so that you don't have to tweak anything to run it. If you run the code on Kaggle website than you don't even have to install any libraries.
To install the libraries just enter the command: pip install "library name" in command prompt or windown powershell.
Do tweak with different hyperparameters to increase the accuracy of the model.
Author: Aaryan Regmi