ZPY: Open Source Synthetic Data for Computer Vision

2021-08-06 · Github 2021 8 · Hugo Ponte, Norman Ponte, Sammie Crowder, Kory Stiger, Steven Pecht, Elena Ponte

Synthetic data presents a unique solution to the huge data requirements of computer vision with deep learning. In this work, we present zpy, an open source framework for creating synthetic data in Python. Built on top of the popular open source 3D toolset Blender, zpy is designed with accessibility and readability in mind. Open source synthetic data toolkits like zpy are the bridge between the more mature tools of the 3D workflow and the machine learning frameworks. We make the case for why open source synthetic data is important to solve issues such as fairness and bias by democratizing access to data. Finally, we explore the effect of different types of domain randomization on synthetic training data by fine tuning a CNN on small synthetic training datasets and testing the model on a holdout test dataset of real images. All code is available on GitHub at http://github.com/ZumoLabs/zpy

📄 PDF Abstract BibTeX

Code (1)

ZumoLabs/zpy

Tasks

Fairness Synthetic Data Generation

Methods 이 논문이 사용한 방법론

Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…

RoIAlign Region of Interest Align, or RoIAlign, is an operation for extracting a small feature map from each RoI in detection and segmentation based tasks. It removes the harsh…

RoIPool 설명 없음

Unity Perception: Generate Synthetic Data for Computer Vision

2021-07-09 · Steve Borkman, Adam Crespi, Saurav Dhakad, Sujoy Ganguly 외

We introduce the Unity Perception package which aims to simplify and accelerate the process of generating synthetic datasets for computer vision tasks by offering an easy-to-use and highly customizable toolset. This open…

2D Object Detection object-detection Object Detection Unity

On The State of Data In Computer Vision: Human Annotations Remain Indispensable for Developing Deep Learning Models

2021-07-31 · Zeyad Emam, Andrew Kondrich, Sasha Harrison, Felix Lau 외

High-quality labeled datasets play a crucial role in fueling the development of machine learning (ML), and in particular the development of deep learning (DL). However, since the emergence of the ImageNet dataset and the…

Continual Learning

NutritionVerse-Synth: An Open Access Synthetically Generated 2D Food Scene Dataset for Dietary Intake Estimation

2023-12-11 · Saeejith Nair, Chi-en Amy Tai, Yuhao Chen, Alexander Wong

Manually tracking nutritional intake via food diaries is error-prone and burdensome. Automated computer vision techniques show promise for dietary monitoring but require large and diverse food image datasets. To address …

Diversity

InfiniteForm: A synthetic, minimal bias dataset for fitness applications

2021-10-04 · Andrew Weitz, Lina Colucci, Sidney Primas, Brinnae Bent

The growing popularity of remote fitness has increased the demand for highly accurate computer vision models that track human poses. However, the best methods still fail in many real-world fitness scenarios, suggesting t…