ResNet34 on CIFAR-10

This repository contains the implementation of a ResNet34 convolutional neural network (CNN) for image classification on the CIFAR-10 dataset. The model is implemented using PyTorch, and training/testing has been done with the CIFAR-10 dataset, which consists of 10 classes of images.

Table of Contents

• Installation
• Usage
• Model Architecture
• Training
• Evaluation
• Predicting on New Images
• Results
• References

Installation

1. Clone this repository:

   git clone https://github.com/devs-mohanraj/implementation-Of-ResNet.git
   
   cd implementation-Of-ResNet

2. Install the required packages:

   pip install -r requirements.txt

3. Download the CIFAR-10 dataset during training, or you can manually download and place it in the data directory.

   ##to download the dataset
   wget https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz 
   
   ##to extract the file 
   tar -xvzf cifar-10-python.tar.gz

Usage

Training the Model

You can train the ResNet34 model on the CIFAR-10 dataset by running:

python train.py

Predicting on new Images

You can visualize the prediction by running the script with an image file, and it will display the image with the predicted label.

python predict.py --image-path /path/to/your/image.png

Model Architecture

The architecture used in this project is based on the ResNet34 model, a 34-layer deep residual network. It uses skip connections and batch normalization layers to avoid vanishing gradients during training. The key components of the model are:

BasicBlock: This forms the building block of the network.

ResNet: The main network class that stacks multiple layers of BasicBlock.

ResNet34(
  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  (layer1): Sequential(...)
  (layer2): Sequential(...)
  (layer3): Sequential(...)
  (layer4): Sequential(...)
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc): Linear(in_features=512, out_features=10, bias=True)
)

the architecture in a tabular representation

the outcomes will be ....

Results

Validation Loss : 0.8831 / ~21.57%

Validation Accuracy : 78.43%

Acknowledgements

the ResNet Original Paper

All you have to know about CIFAR10