There are many pre-trained models which have been trained on millions of images from the ImageNet database with 1000 different classes which include Inception, ResNet, DenseNet to name a few. However, these models have a large number of parameters, which makes them heavy and unsuitable to be used in low computing and powered devices like Mobile phones.MobileNet was developed by Google’s research team specifically for mobiles, which have low memory and power requirements.
This repository contains the details of how to train a lightweight mobilenet for bottle classification. The data is available at my kaggle and can be accessed from https://www.kaggle.com/datasets/zahidbooni/bottles-dataset
The dataset contains 19192 small , 9781l large amd 4675 non-bottles objects respectively. I have uploaded the notebook on this repo , it can also be accessed from kaggle using this https://www.kaggle.com/code/zahidbooni/bottles-classification
a sample from each class can be seen as follows
