To utilize and improve upon pre-trained deep learning models for the classification of Caltech-256 images and achieve a high accuracy final model
To study the dataset and perform Exploratory Data Analysis. To preprocess the dataset with oversampling, under sampling and augmentation techniques. To build and compare the results of hyper parameterization using ResNet18 and ResNet50 models.
The Caltech-256 dataset is a collection of 30,607 images across 257 categories, with a minimum of 80 images per category. The images are of varying sizes. The dataset covers a wide range of objects, such as animals, vehicles, and natural objects.
Image classification involves categorizing images into specific classes or labels. The goal is to use and compare models that provide best accuracy in differentiating between different categories of images. Techniques used pre-trained CNN models such as ResNet18 and ResNet50 and hyper parameter tuning.
