Modern SOTA CNN Architectures

Learning Objectives:

Upon completing this project, you will achieve the following learning objectives:

1. Build Important Blocks for Modern CNNs:

   • Gain practical experience in building key blocks used in modern Convolutional Neural Networks (CNNs).
     • Implement Residual connections to facilitate the training of deep networks.
     • Incorporate Batch Normalization for improved training stability and faster convergence.
     • Utilize Depthwise Separable Convolution for efficient model architecture.

2. Interpret What CNNs Learn:

   • Develop skills in interpreting and visualizing what CNNs learn.
     • Visualize activations to understand the active regions in the network.
     • Explore visualizing filters to comprehend the learned features by individual filters.
     • Generate heatmaps to highlight the crucial areas of focus in the input images.

Files Contained in the Project:

• Modern_CNN_Architectures.ipynb: Jupyter notebook containing the project implementation and experimentation.

Usage:

1. Clone the repository:

git clone https://github.com/Praveen76/Modern-SOTA-CNN-Architectures.git
cd Modern-SOTA-CNN-Architectures

2. Open and explore the Jupyter notebook Modern_CNN_Architectures.ipynb to understand the project implementation.

3. Execute the code cells within the notebook to experiment with building important blocks for modern CNNs and interpreting CNN learning.

4. Gain insights into the usage of residual connections, batch normalization, and depthwise separable convolution in modern CNN architectures.

Feel free to contribute, report issues, or suggest improvements!

Contributing

If you have a Data Science mini-project that you'd like to share, please follow the guidelines in CONTRIBUTING.md.

Code of Conduct

Please adhere to our Code of Conduct in all your interactions with the project.

License

This project is licensed under the MIT License.

Contact

For questions or inquiries, feel free to contact me on Linkedin.

About Me:

I’m a seasoned Data Scientist and founder of TowardsMachineLearning.Org. I've worked on various Machine Learning, NLP, and cutting-edge deep learning frameworks to solve numerous business problems.