Synthetic Data Generation for Enhanced Computer Vision applications: A CAD model and Blender approach
- In today's AI-driven era, computer vision, including autonomous driving, robotics, and healthcare, is prevalent. How-ever, acquiring ample data while managing resources and privacy constraints is challenging. This article proposes a solution: synthetic data generation. We use CAD software to craft intricate 3D models, process them in Blender, and evaluate quality using metrics like Structural Similarity and PSNR (Peak Signal to Noise Ratio). Synthetic data achieves up to 90% similarity with real data and an average PSNR of 21dB. Our method offers a streamlined, dependable ap-proach for enhancing computer vision, especially in object detection, addressing data acquisition challenges.
Author: | Dhanush Gangadharaiah |
---|---|
URN: | urn:nbn:de:bsz:747-opus4-7481 |
Document Type: | Working Paper |
Language: | English |
Date of Publication (online): | 2024/03/15 |
Release Date: | 2024/04/08 |
GND Keyword: | Image Processing; Realistic Data Generation; Structural Similarity Index; Peak Signal to Noise Ratio |
Page Number: | 7 |
Institutes: | Hochschule Ravensburg-Weingarten |
Licence (German): | Creative Commons - CC BY-SA - Namensnennung - Weitergabe unter gleichen Bedingungen 4.0 International |