Backbone of Computer Vision

This page focuses on building convolution layers from scratch so we can understand what is happening under the hood.

• Alexi

Full CNN Code for Image Classification:

import torch
import torch.nn as nn
import torch.nn.functional as F

class SimpleCNN(nn.Module):
    def __init__(self):
        super(SimpleCNN, self).__init__()
        # First convolutional layer
        self.conv1 = nn.Conv2d(in_channels=1, out_channels=32, kernel_size=3, stride=1, padding=1)
        # Pooling layer
        self.pool = nn.MaxPool2d(kernel_size=2, stride=2)
        # Second convolutional layer
        self.conv2 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, stride=1, padding=1)
        # Fully connected layers
        self.fc1 = nn.Linear(64 * 7 * 7, 128)
        self.fc2 = nn.Linear(128, 10)

    def forward(self, x):
        # Apply first convolutional layer followed by ReLU activation and pooling
        x = self.pool(F.relu(self.conv1(x)))
        # Apply second convolutional layer followed by ReLU activation and pooling
        x = self.pool(F.relu(self.conv2(x)))
        
        # Flatten the output for the fully connected layer
        x = x.view(-1, 64 * 7 * 7)
        # Apply first fully connected layer with ReLU
        x = F.relu(self.fc1(x))
        # Apply second fully connected layer (output layer)
        x = self.fc2(x)
        return x

# Create a CNN instance
model = SimpleCNN()

# Example input (batch of 1 image, 1 channel, 28x28 pixels)
example_input = torch.randn(1, 1, 28, 28)

# Get output
output = model(example_input)

Focus of this page:

# First convolutional layer
self.conv1 = nn.Conv2d(
											in_channels=1, 
											out_channels=32, 
											kernel_size=3, 
											stride=1, 
											padding=1
											)
# Pooling layer
self.pool = nn.MaxPool2d(
												kernel_size=2, 
												stride=2
											)
											
# Second convolutional layer
self.conv2 = nn.Conv2d(
											in_channels=32, 
											out_channels=64, 
											kernel_size=3, 
											stride=1, 
											padding=1
											)
        

We will be looking at conv1d to conv3d layers and it’s implementation.

1D Convolution Layer:

• Forward Pass — From Scratch implementation w/ Numpy:
• Forward Pass + Backpropogation — From Scratch implementation w/ Pytorch:

2D Convolution Layer: