• 4 min read
How to Scale a Computer Vision Pilot to Production in Manufacturing
See how to scale a computer vision pilot to production: why 77% stall on integration, not the model, and how to get past it in manufacturing.
Blog
Erik Kokalj
Developer Experience @ Roboflow
Latest Posts by Erik Kokalj
• 4 min read
Making Visual AI Standard Practice in Complex Manufacturing
Learn how to implement visual AI in manufacturing from Florida Crystals' quality lead: trust tests, feedback loops, and pilots that scale.
• 2 min read
Claude Sonnet 5 for Vision: Evaluation and Benchmarks
We ran Claude Sonnet 5 through the Roboflow Vision Evals across 67 real vision prompts. It ties Sonnet 4.6, trails Fable 5, and sits well behind Gemini 3.5 Flash, which still leads the board.
• 5 min read
Turning Computer Vision Into Real‑World Value at Enterprise Scale
Learn why computer vision pilots fail in the real world. Roboflow CEO Joseph Nelson shares a three-step playbook for deploying vision AI in production.
• 4 min read
Human-in-the-Loop Computer Vision: Closing the Gap Between the Model and the Floor
See how human-in-the-loop computer vision closes the gap between model metrics and the floor. Watch operators improve the next model live.
• 6 min read
What Is an eQMS in Manufacturing?
Discover what an eQMS is and how integrating computer vision automates manufacturing quality control, from real-time defect detection to CAPA tracking.
• 7 min read
REST APIs in Manufacturing
Learn how REST APIs connect vision AI, MES, and quality systems on the plant floor. Understand when to use REST versus MQTT and OPC UA for industrial data.
• 2 min read
Claude Fable 5 for Vision: Evaluation and Benchmarks
Claude Fable 5 is a strong reasoning model for visual understand but not a state-of-the-art vision model.
• 6 min read
MQTT in Manufacturing with Vision AI
Discover how MQTT connects computer vision models to PLCs, SCADA, and MES systems on the plant floor. Learn to automate defect detection and safety alerts.
• 6 min read
The Computer Vision Manufacturing Integration Stack
Connect computer vision to your manufacturing stack: edge inference plus PLC, SCADA, MES, and ERP integration over EtherNet/IP, MQTT, and OPC UA.
• 4 min read
Cosmos 3: Evaluation for Vision Use Cases
NVIDIA Cosmos 3 offers zero-shot performance on fixed-camera footage across airports, warehouses, and production lines.
• 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
Serverless GPU Inference Cost Comparison: Roboflow, GCP, AWS, Azure
SUMMARY
Serving a custom RF-DETR XL model on serverless GPU infrastructure produces dramatically different monthly costs depending on the provider and traffic pattern, so this post benchmarks Roboflow Serverless, GCP Cloud Run, AWS SageMaker, and Azure Serverless GPU across three workloads: continuous inference at one request per 10 seconds,
• 4 min read
What Are Anchor Boxes in Object Detection?
Learn what anchor boxes are, and why they are used in object detection.
• 5 min read
Which is the Best Coding Agent for Vision tasks?
SUMMARY
This benchmark pits four coding agents (Claude Code with Opus 4.6, Cursor with Composer 2, Gemini CLI with Gemini 3.1 Pro, and Codex with GPT 5.4) against five computer vision tasks including bird counting with SAM 3, car counting in video and RTSP streams, avocado detection,
• 7 min read
VoTT for Image Annotation and Labeling
A guide on using VoTT to label your own computer vision dataset.
• 7 min read
How to Train a Custom TensorFlow Lite (Now LiteRT) Object Detection Model
Learn how to train a custom TensorFlow Lite object detection model with a custom dataset.
• 7 min read
How to Train YOLOv4 on a Custom Dataset
Want to train YOLOv4? Train a custom object detector without Darknet, a config file, or a GPU: Learn how with Roboflow.
• 10 min read
How to Train an EfficientDet Object Detection Model with a Custom Dataset
Learn how to train an EfficientDet object detection model with a custom dataset.
• 8 min read
How to Train a VGG-16 Image Classification Model on Your Own Dataset
The VGG16 model explained: 16 layers, 3x3 filters, 92.7% ImageNet top-5, plus how to train an image classifier on your own data, no GPU.
• 5 min read
Why and How to Implement Random Rotate Data Augmentation
Learn how to apply a random rotate data augmentation to images for use in training computer vision models.
• 7 min read
EfficientDet for Object Detection
EfficientDet object detection explained: EfficientNet backbone, BiFPN feature fusion, and compound scaling, plus how to train a detector today.
• 7 min read
EfficientDet vs YOLO for Object Detection
EfficientDet vs YOLO compared on architecture, speed, accuracy, and licensing, plus why RF-DETR is the stronger detector to deploy today.
• 9 min read
How to Train an EfficientNet Model with a Custom Dataset
Learn how to train an EfficientNet model with a custom dataset.
Stay Connected