Member-only story
Create YOLO model from scratch
In this tutorial we will be building a YOLO model from scratch. Once completed you will have a good understanding of all object detector components.
Introduction
YOLO is a very popular computer vision model used for object detection and classification. It’s fast and reliable as the name suggest: You Only Look Once.
We’ll build a YOLOv3 style detector from scratch in PyTorch: computing anchors, building dense targets, writing the training loop, and training on Pascal VOC.
This is for readers comfortable with Python and basic PyTorch (tensors, nn.Module, optimizers) who want to understand how YOLO works under the hood.
Prerequisites
To complete this tutorial software programming knowledge is absolutely necessary. A working software development environment is also crucial. It’s assumed that the reader has all the necessary tools configured, for example Visual Studio Code, Anaconda and Python. Familiarity with basic convolutional networks and GPU training is recommended.
