Classifying CIFAR10 Image dataset using a deep CNN.
The project focuses on building a Deep Convolutional Neural Network (CNN) for classifying colored digital images into different categories. Image classification is vital for various applications like self-driving vehicles and weather prediction. Conventional methods such as the Minimum Distance and Maximum Likelihood Classifiers are inferior to CNNs in terms of time and resource consumption. The structure of digital images is discussed, along with Artificial Neural Networks (ANNs) and activation functions like Sigmoid, RELU, and Softmax. Deep learning is explained as a widely used approach in machine learning. The training process involves supervised learning and includes initializing filters and weights, forward and backward propagation, and applying the Gradient Descent algorithm. After training, new images are classified through forward propagation using the trained weights.
- Python
- Tensorflow & Keras
- CUDA to use GPU in training