• 16 min read
How to Train a YOLOv6 Model on a Custom Dataset
Learn how to train a YOLOv6 object detection model on a custom dataset.
Blog
Joseph Nelson
Roboflow cofounder and CEO. On a mission to transform every industry by democratizing computer vision. Previously founded and sold a machine learning company.
Latest Posts by Joseph Nelson
• 10 min read
How to Deploy Computer Vision Models: Best Practices
In this guide, we walk through the fundamentals of deploying vision models and the questions you should evaluate when deciding how to deploy a model.
• 6 min read
Let's Talk About Company Culture
Company culture is a feeling in the air, the social norms of a group, the manifestation of collective goals and beliefs, and the one intangible thing that can make or break a growth stage company. For some, culture becomes a priority too late in the business. It starts to become
• 1 min read
What is PaddlePaddle?
SUMMARY
PaddlePaddle (PArallel Distributed Deep LEarning) is Baidu's open-source deep learning framework, comparable in scope to PyTorch and TensorFlow, providing the core primitives needed to build and train deep learning models. It gained attention in computer vision through PP-YOLO, a model in the YOLO family that
• 1 min read
What is PyTorch?
SUMMARY
PyTorch is Facebook's open-source machine learning framework, originally popularized in academic research and now widely used in production systems. It provides the foundational building blocks for deep learning across computer vision, natural language processing, and audio, and serves as the base for higher-level frameworks like
• 3 min read
We Raised a $20M Series A to Continue Democratizing Computer Vision
Today, Roboflow is excited to announce our $20 million Series A led by Craft Ventures to expand on our mission to give software the sense of sight. Lachy Groom, who co-led the seed with Craft, also participated in the round. Additionally, we welcomed new investment from leaders like Jack
• 3 min read
What is TensorRT?
SUMMARY
TensorRT is NVIDIA's inference optimization framework, built to run machine learning models as fast as possible on NVIDIA GPU hardware via CUDA cores. To use it, you typically train in PyTorch or TensorFlow and then convert your model to TensorRT format. This post covers the CUDA vs
• 1 min read
What is Amazon Rekognition?
SUMMARY
Amazon Rekognition is an off-the-shelf computer vision API that detects and classifies a fixed set of common objects (furniture, plants, and similar generic categories) from images sent via HTTP. It works well for those predefined categories, but if your application needs to detect domain-specific objects not
• 1 min read
What is CoreML?
SUMMARY
CoreML is Apple's on-device machine learning inference framework, designed to run models efficiently on iPhones and other Apple hardware by routing computation through Apple's Neural Engine for lower latency and reduced power draw. Models are stored as .mlmodel files and can be created with
• 1 min read
What is ONNX?
Learn what ONNX is and when it i sused in computer vision.
• 2 min read
Roboflow for Students and Universities
We've seen tremendous interest in Roboflow from the research community. Faculty and students from institutions ranging from California to Malta to Taiwan have been using Roboflow to accelerate their computer vision work.
• 3 min read
Using Computer Vision to Clean the World's Oceans
SUMMARY
Researchers from CSU Monterey Bay, The Ocean Cleanup, and UC San Diego built a computer vision model to detect underwater plastic debris for autonomous underwater vehicles. Using a 3,200-image dataset collected from real California field sites and annotated with Roboflow, the team compared YOLOv4, YOLOv4-tiny, and
• 7 min read
How Atos Uses Computer Vision to Monitor Office Occupancy
SUMMARY
Atos, a digital transformation firm with over 100,000 employees, built a privacy-first office occupancy counter using computer vision and Roboflow, going from initial data collection to a production client demo running on an NVIDIA Jetson Nano in under 60 days. The system connects to existing security camera
• 5 min read
How to Train with Microsoft Azure Custom Vision and Roboflow
SUMMARY
Teams with existing Microsoft Azure credits can route their Roboflow datasets directly into Azure Custom Vision for training by linking API keys once at the workspace level. After that connection is established, any versioned Roboflow dataset (including preprocessing and augmentation steps applied in Roboflow) can be exported to Azure
• 6 min read
How Computer Vision Streamlines Risk Avoidance Workflows in Oil & Gas
SUMMARY
In Canadian oil and gas operations, regulators require companies to identify and notify landowners whose property falls within a proposed pipeline's consultation area, a task that traditionally meant manually digitizing house locations one by one in ArcGIS. This guest post by Douglas Long describes replacing that manual
• 4 min read
Prompt Engineering: The Magic Words to using OpenAI's CLIP
SUMMARY
OpenAI's CLIP model performs zero-shot image classification by matching text prompts to image content, but the exact wording of each prompt has a significant effect on accuracy. This post walks through that sensitivity using a rock, paper, scissors hand-sign dataset, showing how prompts like "
• 3 min read
Partnering with Luxonis and OpenCV for Seamless Deployment to OpenCV AI Kit
SUMMARY
Roboflow and Luxonis have partnered to let users train a custom computer vision model and deploy it to the OAK-1 or OAK-D via a Docker container, eliminating the manual steps of model format conversion to OpenVINO and dependency setup that previously caused silent accuracy regressions. After training
• 3 min read
Image Augmentations for Aerial Datasets
Learn how to apply image augmentations to aerial datasets for use in training computer vision models.
• 3 min read
Our Favorite Computer Vision Courses
SUMMARY
The right computer vision course depends heavily on what you want to build and how much you already know. This guide organizes recommendations by learner type: Andrew Ng's AI for Everyone for non-technical conceptual grounding, fast.ai and Roboflow's own tutorials for practitioners who
• 7 min read
How to Use Roboflow with IBM Visual Recognition (IBM Watson vs Roboflow)
SUMMARY
IBM shut down IBM Visual Recognition in 2021, leaving teams mid-project and needing a path forward. This post walks through two methods for exporting data out of IBM Visual Recognition and IBM Cloud Annotations, then compares what to look for in a replacement platform, covering annotation speed, active
• 4 min read
What Is Embedded Machine Learning?
SUMMARY
Embedded machine learning is the practice of running inference directly on microcontrollers and low-power hardware, such as a Raspberry Pi, NVIDIA Jetson, or Luxonis OAK, rather than sending data to a cloud server. Shrinking model sizes and cheaper, faster hardware have made this practical: inference requires far fewer
• 9 min read
How We Built Paint.wtf, an AI Game with 150,000+ Submissions that Judges Your Art
SUMMARY
Paint.wtf is an online drawing game built by Roboflow and Booste that uses OpenAI's CLIP model to score user submissions against absurdist text prompts, ranking drawings by how closely their image embeddings match the prompt's text embedding. GPT-2 generated the bulk of the
• 4 min read
Using Computer Vision to Help Win $1 Million in Mountain Dew's Big Game Contest
SUMMARY
Mountain Dew's Super Bowl LV ad challenged viewers to count the exact number of bottles in the commercial for a $1 million prize, and this post builds an object detection model to assist with that count. The 30-second video was sampled at three frames per second
• 7 min read
How to Train and Deploy a License Plate Detector to the Luxonis OAK
In this post, we will leverage Roboflow and the Luxonis OAK to train and deploy a custom license plate model to your OAK device.
Stay Connected