README.md

CIFAR-10 Vision Quest: Exploring Image Classification with Deep Learning

Overview

This project focuses on classifying images from the CIFAR-10 dataset using a variety of machine learning models. The models explored include:

• Logistic regression
• Decision trees
• Convolutional neural networks (CNNs)
• Transformers

Training and Evaluation

The models were trained and evaluated to compare their performance using standard metrics such as accuracy. This approach allows for a comprehensive comparison of different machine learning techniques in the context of image classification.

Key Findings

Our results demonstrate the superior effectiveness of deep learning models in handling complex image classification tasks. Specifically:

• CNNs showed exceptional performance
• Transformers also exhibited high effectiveness

These findings highlight the power of advanced neural network architectures in processing and classifying visual data from the CIFAR-10 dataset.

Project Structure

root
├── README.md
├── project_report.pdf
├── EDA_ML_demo.ipynb
├── ResNet
    ├── Resnet18_train.ipynb
    ├── demo.ipynb
    ├── cifar10-resnet18.ckpt
├── CCT
    ├── code            
    │   ├── Configure.py          
    │   ├── DataLoader.py         
    │   ├── ImageUtils.py 
    |   ├── loss.py
    |   ├── main.py 
    |   ├── Model.py
    |   ├── Network.py
    |   ├── Readme.md
    |   ├── requirements.txt           
    └── saved_models
    |   ├── checkpoint.pth
├── multi_modal
    ├── demo.ipynb

Running Exploratory Data Analysis and Machine Learning models

cd root/ Open the python notebook EDA_ML_demo.ipynb to run the file

Running the ResNet model

cd root/ResNet/

--- Training the model ---

Open Resnet18_train.ipynb

--- Testing the model ---

Open demo.ipynb

Notes

1. Ensure that the cifar10-resnet18.ckpt is present while evaluating the model.

Running the Compact Convolutional Transformer model

cd root/CCT/code/

--- Training the model ---

python main.py "train" ./data

--- Testing the model ---

python main.py "test" ./data

Notes

1. Please note that CIFAR-10 data is downloaded into ./data folder for the first time.
2. Ensure the GPU mode is enabled as it takes lot of time in CPU mode.

Running the Multi-modal model

cd root/multi_modal

Open demo.ipynb