GIF_convolutions

Full explanation of the Conv2D : Conv2d: Finally Understand What Happens in the Forward Pass

Make your own gifs of Conv2d forward pass

Download or clone the folder

Requirements

Run in your bash

pip install -r requirements.txt

or

pip install numpy opencv-python scikit-image imageio

Customize your GIF

if you want to customize your GIF : all the arguments are written at the top of the main.py script. You can change with the values you want :

• Height of input channels
• Width of input channels
• Number of input channels
• Number of output channels
• Kernel size
• Padding
• Strides
• Dilations
• Groups

If you also want to change the colors. Go to the pixel.py script at the top of the Pixel class. And if you want to change the padding of the image and the margins between Inputs, Kernels and Outputs, you can refer to line 80 of the conv2d.py script where the renderer.get_full_image() method is called.

Test the impact of the arguments with Pytorch

You will of course need Pytorch for this part. Every script is in the folder run_test/. It is highly recommended to have the GPU drivers. If not, the tests will run on the CPU instead of the GPU.