• 7 min read
How to Use Tiling During Inference
Discover how image tiling improves small object detection. Learn to apply it during training and inference using Roboflow Workflows.
• 12 min read
How to Deploy Computer Vision Models Offline
In this guide, we walk through how to deploy computer vision models offline using Roboflow Inference.
• 6 min read
Gemini 3.5 Flash for Vision: Evaluation and Benchmarks
SUMMARY
Gemini 3.5 Flash, released at Google I/O on May 19, 2026, currently tops the Roboflow Vision Evals leaderboard across 67 real vision prompts covering defect detection, document understanding, object counting, and spatial reasoning. It outperforms Gemini 3.1 Pro on counting and spatial tasks while running roughly
• 5 min read
Integrate Roboflow with OPC UA
SUMMARY
The Roboflow OPC UA Writer Sink block (available on Enterprise plans) lets a Roboflow Workflow push detection results directly to an OPC UA server as live tags, making computer vision outputs visible in SCADA platforms, PLCs, and dashboards alongside other plant sensor data. This guide walks through starting the
• 3 min read
Launch: 300+ OpenRouter models available in Roboflow
SUMMARY
Roboflow's OpenRouter integration adds a single Workflow block that routes to any of 300+ VLMs, including every major commercial model and a broad selection of open-source options, without requiring separate auth, request adapters, or retry logic for each provider. Practical patterns this enables inside Roboflow Workflows
• 5 min read
GPT-5.5: Vision Benchmarks & Use Cases
Explore OpenAI GPT 5.5 vision benchmarks. Learn how its new high-resolution encoder improves document parsing and how to use it in Roboflow Workflows.
• 4 min read
Claude Opus 4.7: Vision Benchmarks & Use Cases
See Claude Opus 4.7 vision benchmarks. Learn how to use its higher-resolution image encoder and document parsing for automated data labeling.
• 4 min read
Inference 1.0: Foundational Infrastructure for Visual Understanding
SUMMARY
Roboflow Inference 1.0 is a modular, multi-backend execution engine built to run vision models at enterprise scale across cloud and edge hardware. The 1.0 release adds automatic backend selection across ONNX, PyTorch, and TensorRT, so the server picks the best runtime for the underlying hardware without
• 4 min read
Launch: Train and Deploy YOLO26 with Roboflow
SUMMARY
Roboflow platform support for YOLO26 covers the full workflow from labeling and training to deployment. The model's edge-optimized architecture, which removes Non-Maximum Suppression for end-to-end predictions and introduces the MuSGD optimizer, can be deployed via Roboflow Inference on CPU or GPU hardware. Dataset
• 6 min read
Launch: Use Segment Anything 3 (SAM 3) with Roboflow
SUMMARY
Roboflow has integrated SAM 3, Meta's foundation model for Promptable Concept Segmentation, across every major part of its platform, covering the Playground for interactive testing, the inference API, edge and private-cloud deployment, Roboflow Workflows for building production pipelines, and the Label Assist annotation tool for accelerating
• 3 min read
Roboflow Named Leader in Image Recognition Software
We’re proud to announce that as of November 2025, Roboflow has been ranked the #1 Image Recognition product on G2.
• 3 min read
NYCerebro: Winning a GPU Signed by Jensen Huang
SUMMARY
NYCerebro is a semantic search tool built in two hours at the Vercel and NVIDIA World's Shortest Hackathon that lets users query New York City's public traffic cameras using natural language. The system uses OpenAI's CLIP model to embed each camera's
• 3 min read
Automate Game Scorekeeping with Vision AI
SUMMARY
This tutorial builds a real-time automated scorekeeping system for cornhole using a custom computer vision model and a Roboflow Workflow. The model detects bags, the board, and the hole from an overhead camera; the Workflow then applies game logic to map detections to zones, calculate the score, and
• 4 min read
Launch: Dedicated Deployments
SUMMARY
Roboflow Dedicated Deployments provision exclusive CPU or GPU servers pre-configured with Roboflow Inference, removing the need to manage hardware, drivers, TLS certificates, or scaling yourself. Two environment types are available: a short-lived development tier billed by the hour for testing and prototyping, and a persistent production tier
• 5 min read
Count Objects on a Conveyor Belt Using Computer Vision
SUMMARY
This tutorial builds an end-to-end system for counting bolts and nuts on a moving conveyor belt, covering every stage from data collection to local deployment. Video frames are annotated using Roboflow's auto-labeling with Grounding DINO, a custom detection model is trained in the cloud,
• 10 min read
How to Use Your GPU in a Docker Container
Learn how to use your GPU in a Docker container and measure upir application's performance.
• 4 min read
How to Draw a Bounding Box with Python
In this post, we discuss how to use the Roboflow supervision library to draw and fill a bounding box in Python.
• 7 min read
What is ByteTrack? A Deep Dive.
SUMMARY
ByteTrack is a multi-object tracking algorithm that improves on earlier methods by using all detections, not just high-confidence ones, to maintain consistent object identities across video frames. Its two-stage data association module first matches high-confidence detections to existing tracklets, then reconciles lower-confidence ones using
• 13 min read
How to Map User Inputs in Video Games with Computer Vision
SUMMARY
This tutorial uses computer vision to track player inputs in the mobile game Brawl Stars by detecting the positions of on-screen joystick and attack buttons frame by frame. A Roboflow object detection model is trained on 8 in-game button classes, a Workflow handles inference, and the Roboflow
• 8 min read
How to Build a Reading Assistant with AI
SUMMARY
This tutorial builds an interactive reading assistant that detects a finger pointing at a word on a page using an object detection model, reads the word using OCR, and then calls GPT-4 to pronounce it aloud, all wired together inside a Roboflow Workflow and run as a live
• 6 min read
How to Deploy Hugging Face Models with Roboflow
SUMMARY
Computer vision model weights hosted on Hugging Face, including YOLOv8 variants for detection, segmentation, classification, and keypoint tasks, can be downloaded with Git-LFS and uploaded directly to Roboflow for edge or private-cloud deployment via Roboflow Inference. This guide walks through each step, from cloning model weights to
• 7 min read
Ball Tracking in Sports with Computer Vision
Ball tracking is crucial for AI systems to analyze sports effectively, but it's challenging due to factors like the ball's small size, high velocity, complex backgrounds, similar-looking objects, and varying lighting. This tutorial will teach you how to overcome these challenges.
• 5 min read
How to Use GPT-4 To Extract Handwritten Text from Images
SUMMARY
This tutorial builds a pipeline that combines a custom object detection model trained in Roboflow with GPT-4 Vision inside a Roboflow Workflow to extract and structure handwritten information from homework booklets. The object detection model localizes handwritten marks, and GPT-4 Vision reads time and name fields, with
• 19 min read
How to Monitor Productivity with Eye Tracking
SUMMARY
This tutorial builds a real-time productivity tracker that uses a custom object detection model to monitor eye position, pupil movement, face presence, and phone usage via webcam, then logs distraction events over time. The model is trained in Roboflow on roughly 125 labeled images and deployed locally, where
Stay Connected