This project was completed as a part of the Honors portion of the Convolutional Neural Networks Course on Coursera.
Special thanks to DeepLearning.AI and the Coursera platform for providing the course materials, guidance, and learning resources that made this project possible.
The main objective of this project is to build a deep convolutional neural network using Residual Networks (ResNets) for image classification tasks.
Traditional deep neural networks often suffer from vanishing gradients, which makes training very deep models difficult. Residual Networks (ResNets) solve this issue by introducing skip connections, allowing gradients to flow directly through the network and enabling the training of significantly deeper architectures.
In this project, we:
- Implement the core building blocks of ResNet
- Construct a deep neural network architecture
- Train the model for image classification tasks
- Utilize Keras for building and training the neural network
- Deep Convolutional Neural Networks (CNN)
- Residual Learning
- Skip Connections
- Identity Blocks
- Convolutional Blocks
- Image Classification
- Model Training with Keras
The project implements the fundamental ResNet architecture, which consists of:
-
Identity Blocks
Allows the input to bypass certain layers through skip connections. -
Convolutional Blocks
Adjusts the dimensions of inputs while applying convolution operations. -
Skip Connections
Directly connects earlier layers to later layers to stabilize gradient flow and improve training performance.
These components enable the network to learn deeper and more complex representations.
- Python
- TensorFlow / Keras
- NumPy
- Jupyter Notebook
Deep convolutional networks like ResNet are widely used in:
- Image classification
- Object detection
- Medical image analysis
- Autonomous driving systems
- Facial recognition
- Computer vision research
This project is based on coursework from the Convolutional Neural Networks specialization provided by DeepLearning.AI on Coursera.
Their educational resources and practical assignments played a crucial role in developing the understanding required to complete this project.
HOSEN ARAFAT
Software Engineer, China
GitHub: https://github.com/arafathosense
Researcher: Artificial Intelligence, Image Computing, Image Processing, Machine Learning, Deep Learning, Computer Vision