- Category: Data Science
- Client: University
- Project date: Mar-2021 to May-2021
- Project URL : CIFAR 10 Classification
Summary: The goal of the project was to develop 2 models one of deep learning and one of machine learning, in order to compare the performance of 2 models and approaches on CIFAR-10 dataset which consists of 60000 images and of more than 10 classes
Tools: Jupyter Notebook, Python, Tensorflow, Keras, Pandas, MatplotLib and Sklearn
Duties: It was a team project, where I took over the responsibility of feature engineering, where analysed appropriate method for image classification and extracting features. I used HOG descriptors, a traditional feature engineering and spatial/probability seperation methods to get better representation of the edges identified in images.
Outcome: Out of 2 models deep learning performed better than machine learning, where deep learning achieved almost 85% of accuracy. Also, deep learning has automatic feature extraction, lower training time and more tuning control.
Result: Obtained Distinction score for this project.