The project presents a walkthrough of a jupyter notebook to classify images of nature scenes of 6 different categories with Transfer learning in PyTorch using the Intel Image Classification dataset. The project was built as the final assignment course project for the Deep Learning with PyTorch - Zero to GANS course offered by FreeCodeCamp and Jovian.
The dataset was initially created by Intel for an image classification contest. It is an expansive image dataset consisting of approximately 25,000 images. The images are divided into the following categories: buildings, forest, glacier, mountain, sea, and street. Our task is to train a model using PyTorch to classify the image as one of the 6 categories.
The project work is organised as follows:
- Collecting Dataset
- Defining Problem Statement
- Data Exploration and Cleaning
- Modeling
- Evaluation and Prediction
- Improvement
- Conclusion and Summary
The final model achieved an accuracy of about 93% which can still be improved.
- Dataset - https://www.kaggle.com/puneet6060/intel-image-classification
- Jovian Notebook - https://jovian.ai/lavanyask03/course-project