Machine Learning Model for Flower Image Classification on a Tensor Processing Unit

Identification and categorization of flower images are active research problems in the field of Computer Vision. In the last decade, this problem has been tackled by performing machine learning prediction on the basis of extracted features such as colour or texture extraction. In this project, a novel approach for solving this problem is introduced by integrating and ensembling two efficiently scaled Deep Conventional Neural Network models - EfficientNet and DenseNet. The training experiment would be performed using large number of images from multiple public datasets on multiple complex deep neural network models. To optimize the computational resource and efficiency, the experiment would be run on Tensor Processing Unit (TPU) hardware environment and the efficacy of the same would be assesed in terms of computational power and speed.