The article discusses the creation of animations and instructional videos about neural networks, which are available on Patreon and YouTube. It covers various aspects of convolution in neural networks, including the basic algorithm, padding, stride, and groups, depthwise, and depthwise-separable convolution. Each topic is accompanied by a companion YouTube video for a more detailed explanation.
The article also introduces the concept of pixel shuffle, which is a method of changing resolution with style. It further breaks down pixel shuffle into 2x2 and 3x2 block sizes. The content is licensed under the MIT License.
Key takeaways:
The article provides animations and instructional videos about neural networks, with a focus on convolution.
It discusses the basic algorithm of convolution in neural networks with a 3x3 kernel size, padding, and stride.
The article also covers more advanced topics such as groups, depthwise, and depthwise-separable convolution in neural networks.
It also explores the concept of pixel shuffle, with a focus on 2x2 and 3x3 block sizes.
