Template Credit: Adapted from a template made available by Dr. Jason Brownlee of Machine Learning Mastery.
SUMMARY: This project aims to construct a predictive model using a TensorFlow convolutional neural network (CNN) and document the end-to-end steps using a template. The Fruits-360 Images dataset is a multi-class classification situation where we attempt to predict one of several (for this dataset 131) possible outcomes.
INTRODUCTION: This dataset contains 131 fruits and vegetables with 67692 training images and 22688 test images. All images have a resolution of 100 X 100 X 3 colors in the JPEG format. Each dataset includes 131 subdirectories, one for each type of fruit or vegetable.
From iteration Take1, we constructed a CNN model based on the InceptionV3 architecture to predict the fruit or vegetable name for the image.
In this Take2 iteration, we will construct a CNN model based on the ResNet50 architecture to predict the fruit or vegetable name for the image.
ANALYSIS: From iteration Take1, the InceptionV3 model’s performance achieved an accuracy score of 99.46% after five epochs using the validation dataset. The same model processed the test dataset with an accuracy score of 98.03%.
In this Take2 iteration, the ResNet50 model’s performance achieved an accuracy score of 99.91% after five epochs using the validation dataset. The same model processed the test dataset with an accuracy score of 99.48%.
CONCLUSION: In this iteration, the ResNet50-based CNN model appeared to be suitable for modeling this dataset. We should consider experimenting with TensorFlow for further modeling.
Dataset Used: Fruits-360: A dataset of images containing fruits and vegetables.
Dataset ML Model: Multi-class image classification with numerical attributes
Dataset Reference: https://github.com/Horea94/Fruit-Images-Dataset
One potential source of performance benchmarks: https://www.researchgate.net/publication/321475443_Fruit_recognition_from_images_using_deep_learning
The HTML formatted report can be found here on GitHub.