Dataset Management

2 Feb 2022 • 4 min read

Launch: Collaborative Annotations

SUMMARY Roboflow Annotate now supports team-scale annotation workflows with a board view that shows image status at a glance, job assignment to divide work across team members, reviewer notifications, annotator instructions, and a review mode where a second person can approve or reject labeled images before they enter the

30 Sep 2021 • 2 min read

Roboflow and Ultralytics Partner to Grow Computer Vision Community

SUMMARY Roboflow has become the official dataset management and annotation tool for Ultralytics, with a formal licensing agreement that lets Roboflow users train, deploy, and commercialize YOLO models (including YOLOv8 and YOLO11) through the Roboflow platform without running into AGPL-3.0 licensing conflicts. The partnership also deepens the integration

16 Jul 2021 • 4 min read

5 Reasons to not Fully Outsource Labeling

So you're working on building a machine learning model, and you have hit the realization that you will need to annotate a lot of data to build a performant model. In the machine learning meta today, you will be bombarded with services offering to fully outsource your labeling woes.

20 Jun 2021 • 5 min read

For the People, By the People

SUMMARY Computer vision models are only as representative as the datasets behind them, and gaps in that data encode real-world bias into deployed systems. This post examines what representative training data means across dimensions like skin tone, accessibility, religion, and gender, drawing on examples from hand detection to pedestrian

14 Jun 2021 • 7 min read

Building vs. Buying a Computer Vision Platform

SUMMARY Building a computer vision pipeline in-house requires stitching together image upload, annotation, dataset versioning, training, deployment, and active learning, and the result is typically fragile, hard to maintain, and difficult to debug when components break. This post examines seven recurring problems teams encounter when building their own infrastructure,

24 May 2021 • 3 min read

What is a Cutout Augmentation and When Can it Help?

SUMMARY Cutout augmentation, introduced in the YOLOv4 paper, randomly masks a square region of a training image to force the model to use broader image context rather than relying on a single dominant feature. It is most useful when objects in your dataset overlap or occlude one another (such as

3 May 2021 • 8 min read

The power of image augmentation: an experiment

SUMMARY Image augmentation increases effective training set size by applying random transformations to existing images, but knowing how much to augment and which transforms to apply depends on your dataset. This experiment tests three variables across three public datasets of different sizes (26, 196, and 665 images of packages, raccoons,

26 Apr 2021 • 3 min read

Image Augmentations for Aerial Datasets

Learn how to apply image augmentations to aerial datasets for use in training computer vision models.

26 Jan 2021 • 11 min read

Using Unity Perception to train an object detection model with synthetically generated images

SUMMARY Unity's Perception package can generate perfectly annotated synthetic images using only software, giving computer vision projects a way to bootstrap training data before real-world images are available. This tutorial walks through setting up Unity 2019.4 LTS with the Perception package, generating a labeled synthetic dataset,

6 Jan 2021 • 4 min read

Using Public Datasets to Improve your Computer Vision Models

SUMMARY When a model is underperforming and more field collection is impractical, merging your dataset with publicly available annotated data is often the fastest way to improve accuracy. This post demonstrates the approach using COCO and OpenImages car images, where doubling the training set by adding 1,000 OpenImages examples

25 Dec 2020 • 3 min read

How To Annotate Images with Your Team Using Roboflow

SUMMARY Splitting annotation work across a team in Roboflow involves creating one dataset per annotator, sharing each dataset with the full team so everyone can see examples from colleagues, and then merging the individually labeled datasets into a single combined dataset once all images are labeled. This workflow keeps annotation

8 Dec 2020 • 2 min read

Introducing Roboflow Annotate

SUMMARY Roboflow Annotate is a self-serve image annotation tool built directly into the Roboflow platform, removing the friction of managing a separate labeling tool outside your pipeline. The integrated design supports team collaboration from a single source of truth and makes it straightforward to jump in and out of

30 Nov 2020 • 6 min read

How to Run Jupyter Notebooks on an Apple M1 Mac

Learn how to run Jupyter Notebooks on Apple M1 Macbooks.

28 Oct 2020 • 2 min read

Train Test Split Guide and Overview

SUMMARY Splitting a dataset into training, validation, and test sets is a core practice for detecting overfitting and ensuring a computer vision model generalizes beyond its training examples. Roboflow defaults to a 70/20/10 split at upload time but gives you direct control: you can assign individual image batches

12 Oct 2020 • 2 min read

Introducing an Improved Shear Augmentation

Today, we introduce a new and improved shear augmentation. We'll walk through some details on the change, as well as some intuition and results backing up our reasoning.

27 Sep 2020 • 3 min read

Introducing Grayscale and Hue/Saturation Augmentations

SUMMARY Roboflow has added three color-based augmentation options, grayscale, hue, and saturation, to help models generalize to objects that appear under different lighting conditions or in varied color environments. Grayscale augmentation randomly converts a configurable percentage of training images to single-channel output, reducing the model's reliance

25 Sep 2020 • 4 min read

Getting Started with VGG Image Annotator for Object Detection Tutorial

SUMMARY VGG Image Annotator (VIA) is a free, browser-based labeling tool built by Oxford University researchers that supports bounding box and polygon annotations for object detection datasets. This tutorial covers the full workflow: creating a project, loading images, drawing annotations, assigning class labels, and exporting data, then converting the

4 Sep 2020 • 6 min read

Train, Validation, Test Split for Machine Learning

At Roboflow, we often get asked, what is the train, validation, test split and why do I need it? The motivation is quite simple: you should separate you data into train, validation, and test splits to prevent your model from overfitting and to accurately evaluate your model.

19 Jun 2020 • 5 min read

How to Convert Annotations from PASCAL VOC to YOLO Darknet

SUMMARY PASCAL VOC stores bounding box annotations as XML files with absolute pixel coordinates, while YOLO Darknet uses plain text files with class index and normalized center coordinates. Converting between them means restructuring both the file format and the coordinate representation. This post walks through both formats with concrete examples

13 May 2020 • 7 min read

Data Augmentation in YOLOv4

Learn how data augmentation is used in training YOLOv4 computer vision models.

29 Apr 2020 • 4 min read

Breaking Down Roboflow's Health Check Dimension Insights

SUMMARY Resizing images for computer vision is not a simple stretch-to-fill operation: aspect ratio distortion, small image outliers, and extreme width-to-height ratios all affect what a model can learn. Roboflow's Dimension Insights tool, part of the Dataset Health Check, visualizes every image in a

30 Mar 2020 • 1 min read

Introducing Image Preprocessing and Augmentation Previews

SUMMARY Choosing augmentation settings without previewing the result means iterating blind, generating dataset versions only to discover the brightness shift was too aggressive or the rotation too extreme. Roboflow added preprocessing and augmentation previews so you can hover over an image in the dataset builder and see exactly how each

18 Mar 2020 • 3 min read

Introducing Bounding Box Level Augmentations

SUMMARY Standard image augmentation applies transforms to the entire frame, but bounding box level augmentation applies transforms only to the pixels inside each labeled bounding box, letting you vary brightness, blur, or other properties of a target object independently from its background. This technique, drawn from a 2019 Google research