Labeling Guide: Object Detection
Annotating your images on Roboflow. How to create "ground truth" annotations that lead to better "model predictions." Is the issue with your mAP, precision, and recall "bad data"? Or is the issue, "bad annotations"? This is an important distinction to make. Yes, annotations are painful...but they also have a heavy bearing on the final outcome of you...
Training Your First Model: Roboflow Train
What is Roboflow Train? After labeling your images with bounding boxes, polygons, bounding boxes and polygons, and/or Smart Polygons, and generating a dataset version, you're ready to train a model! Starting Your First Project | Workspaces, Projects, and Versions Roboflow offers an AutoML product called Roboflow Train. It is the easiest way to tra...
Model Training: YOLOv5 Video Tutorial
Train a Custom YOLOv5 Model After generating a version of your labeled object detection dataset, we are ready to train a model. Here is a tutorial on training a YOLOv5 Model with the help of Roboflow's python package for easy data management. Exporting Your Dataset Once you copy your dataset export code, continue on with the video tutorial to train...
Run YOLOv5 on Your Webcam
Local Inference with YOLOv5 on a Webcam Clone the Ultralytics YOLOv5 Github repository to your computer and run detect.py after saving the model weights from your custom YOLOv5 model training. Make a note of the directory (folder) in which you clone the repository. Save the file path, if you can, for easy access later. GitHub commands cheatsheet Inc...
Model Training: YOLOS Video Tutorial
Train a Custom YOLOS Transformer Model This tutorial highlights the combination of a suite of best-in-class machine learning tools - Hugging Face Transformers, Roboflow, Weights and Biases, and AWS Sagemaker Serverless - to create a YOLOS object detection labeling, training, and deployment pipeline. Roboflow Model Library (Model Zoo): YOLOS The YOLO...
Roboflow Glossary
This is a auto-generated Article of all your definitions within the glossary. Glossary This is a auto-generated Article of all your definitions within the glossary. All A B C D E F G H I J K L M N O P S T V W X Y Accuracy The proportion of "correct" vs "incorrect" predictions a model makes. Common in classification models that have a single correct ...
Workspaces, Projects, and Versions
Workspaces All Roboflow projects belong to a workspace. When you sign up for a Roboflow account, we automatically create an initial workspace for you. Workspaces are how you can manage who has access to what projects and what features are available for projects inside the workspace. Workspaces contain Projects which contain Versions that have train...
How are duplicate images counted in my usage?
If an image's content is the exact same, it is only counted towards your usage once. For example, if you merge two datasets, the merged dataset and the two original datasets exist in your account. However, there is no charge for the merge, because the images are duplicates. Merging Projects or Datasets Duplicating Projects or Datasets Roboflow uti...
How do I delete a dataset?
Hover over and click on the 3 horizontal dots next to the dataset you wish to delete. Select "Delete" in the upper right-hand corner. Enter the dataset ID (displayed to you in red). Confirm deletion NOTE : Dataset/project deletions are irreversible !! Do not press "Delete Project" until you have saved a raw image version of your dataset if you would...
Can I edit the "Train, Validation, Test" split after upload?
Yes! We've recently added this feature. When generating a new version of your dataset, click on "Step 2" (Train/Test Split): Roboflow dataset generation UI Next, select "Rebalance" Roboflow train/validation/testing split Rebalancing UI Select your new desired Train/Validation/Test split by adjusting the sliders to their desired positions: Roboflow t...
Model Training: YOLOv4-tiny Video Tutorial
Train a Custom YOLOv4-tiny Model After generating a version of your labeled object detection dataset, we are ready to train a model. Here is a tutorial on training a YOLOv4-tiny Model with the help of Roboflow's python package for easy data management. Exporting Your Dataset There are 4 Ways to Export Your Dataset on Roboflow: (1) Download to a Zip ...
Model Training: YOLOv5-OBB
Train a Custom YOLOv5-OBB Model What is YOLOv5-OBB? YOLOv5 for Oriented Object Detection (Roboflow Blog) Oriented bounding boxes (OBB) are bounding boxes rotated to better fit the objects represented on an angle. Take a pill detection dataset for example. Using YOLOv5-OBB (or YOLOv5-obb) we are able to detect objects that are rotated on a given fram...
Need help with your project?
Having Trouble with Your Project? For Roboflow Public / Community plans: Please visit our bug tracker after searching our forum to learn if its a bug, or potentially miscommunication of how a feature works. Bug Report format Have a bug report for a more urgent matter? Please send an email to help@roboflow.com outlining your request and including det...
Python Package for OAK Deployment
Deploying to OAK Devices with Roboflow's Python Package OAK: OpenCV AI Kit The Luxonis OAK (OpenCV AI Kit) won an award as the 2022 Edge AI and Vision Product of the Year in the Cameras and Sensors category. Follow along with our Quick Start Guide if you haven't yet trained a model with Roboflow Train. If you'd like to access Accurate Train to train...
Deployment: Using the Roboflow Inference API
Model deployment can be a tedious process. Roboflow offers a number of easy deployment options, with model inference code snippets available for command line (cURL), Python, Javascript, Swift, Java, .NET, and more. Roboflow's Hosted Inference API Roboflow's deployment options available after training Your browser does not support HTML5 video. Infere...
Exporting Your Dataset
There are 4 Ways to Export Your Dataset on Roboflow: (1) Download to a Zip Folder (2) Download with our python package - this is the most common method, as it allows you to programmatically download your images and use them to train a model with our model zoo. (3) Download with a curl command (4) Download by Raw URL Your browser does not support HTM...
Dataset Health Check: Improving Your Dataset
Better models start with better datasets. In this video, we walk through common tips and tricks to improve your dataset, including using Roboflow Train and the Dataset Health Check to get there. We walk through a public dataset -- Bike Helmet Detection -- provided via Syad Salman Reza on Roboflow Universe....
Dataset Upload: Roboflow Data Types
How to successfully upload your images, or videos, into Roboflow. Data can currently be added to Object Detection, Single-Label Classification, Multi-Label Classification, Instance Segmentation, and Semantic Segmentation projects (semantic segmentation: upload of masks is not yet supported, only annotations and images - export of masks is supported)...
Merging Projects or Datasets
We've added a feature to merge datasets (with annotations) together in just a few clicks. Hover your mouse on the three dots on the right side of one of the datasets you wish to include the merged dataset. Click the highlighted button (three horizontal dots), and select "Merge Datasets." Select the datasets you wish to merge by clicking the checkbox...
Preparing for Roboflow Universe
How to create your dataset and update your ReadMe ahead of inclusion in Roboflow Universe's Featured section . Roboflow Universe is a great place to receive project feedback and recognition from Roboflow users, and the Roboflow team. We like to select the best projects to highlight on Roboflow Universe's "Featured" section and in our weekly blog an...
Sharing a Workspace with Roboflow Support
You now invite the help@roboflow.com account to your workspace to share your data with the Roboflow team. This will allow our team to access the projects in your workspace and help with debugging and/or issue resolution. To do this, go to your Workspace Settings: Roboflow Workspace homepage UI. Accessing Workspace Settings. "Workspace Settings" can...
A Guide for Model Production
Glossary of Common Computer Vision Terms Key Metrics mAP, Precision and Recall scores are provided for all models trained with Roboflow Train, and through the notebooks in the Roboflow Model Zoo. mAP "When evaluating an object detection model in computer vision, mean average precision is the most commonly cited metric for assessing performance." Wha...
Implementing Active Learning
"Active learning is a machine learning training strategy that enables an algorithm to proactively identify subsets of training data that may most efficiently improve performance. More simply – active learning is a strategy for identifying which specific examples in our training data can best improve model performance. It's a subset human-in-the-loop...
Deleting Annotation Batches
Deleting Annotation Batches on Roboflow You can delete annotation batches for images within your project: Removing Annotation Batches: Images Uploaded to the Project, but Unassigned for Labeling Removing Annotation Batches: Images Uploaded to the Project, and Within the Annotation Queue Removing Annotation Batches: Images are Labeled, Approved, and ...
Creating a Sample App with Universe Projects
Roboflow Universe datasets can be used for project inspiration (use the search bar to look for any terms of interest!), gathering more data for a custom project, and even testing inference on trained Universe models. Brad Dwyer, Roboflow CTO, created an example blackjack strategy web application that utilizes the Augmented Startups Playing Cards dat...
How do I change my workspace from private to public?
Currently, Roboflow has to do this manually for you in our backend before you can transfer a dataset to a public workspace. If you need a workspace change, please add Roboflow Support to your workspace send your request to help@roboflow.com, and include: Workspace name Email that you signed up with Project name(s) for transfer As an added step, plea...
The Labeling Interface
Be sure that you've uploaded a dataset before you get started: Dataset Upload: Roboflow Data Types On the right-hand side of the labeling interface, you will find the Toolbar. It contains the following options: Drag/Select(represented by a hand icon) allows you to select, edit, and drag individual annotations. Single-click an existing bounding box...
Starting Your First Project
Start your first Object Detection project on Roboflow . Dataset Upload --> Dataset Generation & Model Training with Roboflow Train. Plus: Quick insight into Dataset Health Check, selecting preprocessing & augmentation steps, exporting in different annotation formats, and deploying your model to run inference. Topics Covered: Creating Y...
How to Use Roboflow Universe
How to Use Roboflow Universe You're likely on Roboflow Universe to search or explore datasets for computer and machine vision or Digital Transformation (DX); find new data to add to your own project or test and deploy models from Universe in your own applications. Roboflow Universe Overview | Roboflow Universe Announcement Fo...
Dataset Export: YOLOv6 Model Training
MT-YOLOv6, or "YOLOv6," is a single-stage object detection framework dedicated to industrial applications, with hardware-friendly efficient design and high performance. The new object detection model is from 美团 (Meituan Dianping). YOLOv6 benchmarking Background: YOLOv6-nano achieves 35.0 mAP on COCO val2017 dataset with 1242 FPS on T4 using TensorRT...
Roboflow Train: Understanding Training Graphs
Accessing Your Training Graph To access your training graph, you'll first need to train a model with a Roboflow Train credit. All new workspaces come with 3 free Roboflow Train credits. Choosing Between Computer Vision Model Sizes - includes insights on the FAST model vs. ACCURATE model Quick Start Guide Models trained with Roboflow Train give you...