Learn computer vision
Explore our resources that cover all you need to know to plan, label data for, train, deploy, and improve computer vision projects. Roboflow Learn is split into six parts, starting with computer vision foundations and ending with deployment and advanced concepts.
Computer vision foundations
Learn what computer vision is, what you can use computer vision for, and how to train and deploy your first model.
What is computer vision (opens in a new tab)
A good understanding of computer vision without a strong technical background, and the steps needed to solve a computer vision problem. Blog, 12 min read.
Computer vision model types (opens in a new tab)
The spectrum of classification, object detection, and segmentation models, from the Roboflow team. Video, 10:20.
Getting started with Roboflow (opens in a new tab)
Roboflow eliminates boilerplate code when building object detection models. Blog, 17 min read.
Scoping computer vision projects (opens in a new tab)
Choosing the right problem to solve first is half the battle: learnings from observing over 20,000 users' projects. Video, 13:30.
Image labeling and dataset processing
Accurately labeled and prepared data can make or break a vision project. Learn how to label data effectively and add augmentations and preprocessing steps to enhance model performance.
How to label images in Roboflow (opens in a new tab)
An overview of the labeling interface for Roboflow's annotation tool, including shortcut keys. Documentation.
Image labeling best practices (opens in a new tab)
What image labeling is and how to effectively label images for training computer vision models. Blog, 6 min read.
Image preprocessing and augmentation best practices (opens in a new tab)
Understanding preprocessing and augmentation options is essential to making the most of your training data. Blog, 8 min read.
Generating versions (opens in a new tab)
Create a dataset version for use in training a model. Documentation.
Model training
Learn how to train and evaluate a computer vision model using Roboflow Train, or in a notebook.
Train a model in Roboflow (opens in a new tab)
Train a model using state-of-the-art technology in the Roboflow dashboard. Documentation.
Evaluate Roboflow model performance (opens in a new tab)
How to evaluate computer vision models hosted on Roboflow using the in-app model evaluation tool. Blog, 5 min read.
Deploying your model
Learn how to run trained computer vision models in the cloud on images, or on your own device on images and videos.
Deployment best practices (opens in a new tab)
The fundamentals of deploying vision models and the questions to evaluate when deciding how to deploy. Blog, 10 min read.
Deploy models with Roboflow Inference on an image (opens in a new tab)
Run inference on object detection, classification, and segmentation models using Inference. Documentation.
Deploy models with Roboflow Inference on a video or stream (opens in a new tab)
Run models on webcam stream frames, RTSP stream frames, and video frames with Inference. Documentation.
Create a vision application with Workflows (opens in a new tab)
Learn how to identify solar panel locations with computer vision. Blog, 5 min read.
What others have built
Explore examples of projects that solve problems using computer vision.
Case studies (opens in a new tab)
Customer stories on the Roboflow Blog.
Open source projects on Universe (opens in a new tab)
Browse datasets and models on Roboflow Universe.
Logistics use cases with Roboflow (opens in a new tab)
Logistics stories on the Roboflow Blog.
Manufacturing use cases with Roboflow (opens in a new tab)
Manufacturing stories on the Roboflow Blog.
Advanced learning
Read our advanced guides to dive deep into specific computer vision topics.
Best cameras for computer vision (opens in a new tab)
Three cameras we recommend for capturing image frames for computer vision systems. Blog, 7 min read.
How to fine-tune Florence-2 for object detection (opens in a new tab)
Fine-tune Florence-2 on object detection datasets to improve performance for your use case. Blog, 12 min read.
Detect small objects with SAHI (opens in a new tab)
Use the SAHI implementation in the supervision Python package to detect small objects. Blog, 4 min read.
What is Segment Anything 2 (SAM 2)? (opens in a new tab)
Meta AI's Segment Anything 2 model and how to use it for image and grounded image segmentation. Blog, 7 min read.
What are precision, recall, and mAP? (opens in a new tab)
What mean average precision is and how it is calculated. Blog, 10 min read.
Frequently asked questions
You can build a computer vision algorithm in about a day with a tool like Roboflow, which handles the technical back-end and empowers you to focus more on solving a particular problem with computer vision. It takes a few days to learn about the different types of problems you can solve with computer vision and a few weeks to learn more fundamentals like improving model performance and deployment.
With software like Roboflow, you can build a computer vision model without any prior computer vision experience. Previously, learning computer vision involved an extensive investment of time and computing resources. Over the last few years, there have been advances in the field to make the technology more approachable. Now, you can use tools like Roboflow to build models hands-on without minimal to no code, which makes the learning process easier.