CNN Architectures Benchmark

Table of Contents

• 📖Project Description
• 🧑‍💼Business Use Cases
• 📁Data Set Explanation
• 📊Project Evaluation Metrics
• 🚩How to Approach this Project

📖Project Description

The goal of this project is to compare the performance of different CNN architectures on various datasets. Specifically, we will evaluate LeNet-5, AlexNet, GoogLeNet, VGGNet, ResNet, Xception, and SENet on MNIST, FMNIST, and CIFAR-10 datasets. The comparison will be based on metrics such as loss curves, accuracy, precision, recall, and F1-score. Comparison of CNN architectures (LeNet-5, AlexNet, GoogLeNet, VGGNet, ResNet, Xception, SENet) on MNIST, FMNIST, and CIFAR-10 datasets. Evaluates performance using loss curves, accuracy, precision, recall, and F1-score. Implemented with TensorFlow and PyTorch.

🧑‍💼Business Use Cases

The insights from this project can be applied in various business scenarios, including:

• Choosing the appropriate CNN architecture for specific computer vision tasks.
• Improving model performance by understanding the impact of dataset characteristics.
• Optimizing resource allocation by selecting models that offer the best trade-off between performance and computational cost.
• Identifying potential trade-offs between model complexity and performance.
• Understanding the impact of dataset characteristics on model performance.

📁Data Set Explanation

The datasets used in this project are:

• MNIST: Handwritten digits dataset consisting of 60,000 training images and 10,000 testing images. Each image is 28x28 pixels in grayscale.
• FMNIST: Fashion MNIST dataset consisting of 60,000 training images and 10,000 testing images of fashion products. Each image is 28x28 pixels in grayscale.
• CIFAR-10: Dataset consisting of 60,000 32x32 color images in 10 classes, with 50,000 training images and 10,000 testing images.

The datasets are chosen to cover a variety of image classification tasks:

• MNIST and FMNIST provide simpler tasks with grayscale images, allowing for the evaluation of basic image recognition capabilities.
• CIFAR-10 offers a more complex task with color images, testing the models 'abilities to handle more detailed and varied data.

📊Project Evaluation Metrics

The success and effectiveness of the project will be evaluated using the following metrics: -

• Accuracy: The proportion of correct predictions out of the total predictions made.
• Precision: The proportion of true positive predictions out of all positive predictions made.
• Recall: The proportion of true positive predictions out of all actual positives.
• F1-score: The harmonic mean of precision and recall.
• Loss: The value of the loss function during training and testing.

🚩How to Approach this Project

• To understand the project, check out the Approach File.

• You can download all the dependencies by running the requirements.txt file using the following command:

  pip install -r requirements.txt

• Also check out the Final Report for more details about the outcome of the project.