Domain-Adaptation

Extension to YOLOv3 person-detection model by incorporating domain adaptation on 3D synthetic images, allowing for synthetic-to-real translation tasks using Cycle-consistent Generative Adversarial Networks (CycleGAN) for one-to-one mapping framework.

A PyTorch implementation of (Bayesian) CycleGAN inspired by Zhu et al. (2017) and You et al. (2020).

Version: 1.1 (2021)

---

Introduction

introduction 2

Prerequisites

• python 3.8
• PyTorch
• NVIDIA GPU + CUDA CuDNN

Getting Started

Test Results & Performances

Custom Datasets for YOLOv3 Blender Training

Acknowledgements

Code is inspired by .