Computer Vision

8 Aug 2024 • 8 min read

What is Optical Character Verification? A Comprehensive Guide

SUMMARY Optical Character Verification (OCV) reads printed text on product packaging using OCR and then compares the result against known reference data to catch errors in expiration dates, batch numbers, barcodes, and lot codes before products leave the line. In consumer packaged goods manufacturing, OCV automates a check that would

8 Aug 2024 • 8 min read

Camera Calibration in Sports with Keypoints

SUMMARY Accurate player tracking in sports footage requires mapping pixel coordinates from a moving camera to real-world field positions, and this tutorial covers how to do that by training a YOLOv8 keypoint detection model to recognize characteristic landmarks on a soccer pitch, then using those detections to compute a

7 Aug 2024 • 8 min read

How to Extract Text From Images

SUMMARY Optical Character Recognition (OCR) converts static image content in JPG, PNG, or PDF formats into machine-readable text, enabling automation of data entry, accessibility improvements, translation workflows, and document analysis. Roboflow's OCR API is one practical tool for this, offering a programmatic way to extract text at

6 Aug 2024 • 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.

2 Aug 2024 • 3 min read

How to Import Hugging Face Datasets to Roboflow

Learn how to import a Hugging Face dataset into Roboflow for labeling, training, and deployment.

1 Aug 2024 • 7 min read

How to Use SAM 2 for Video Segmentation

SUMMARY Segment Anything Model 2 (SAM 2) is a unified image and video segmentation model from Meta that achieves 3 times fewer required interactions than prior approaches and runs 6 times faster than the original SAM. This walkthrough covers how to load and configure SAM 2 (available in four sizes

30 Jul 2024 • 8 min read

What is Segment Anything 2 (SAM 2)?

Learn about Meta AI's new Segment Anything 2 model and how you can use it for image and grounded image segmentation.

19 Jul 2024 • 5 min read

Evaluating 2024 Euro Cup and COPA America Cup Jersey Color Accessibility

Read how we built a system to evaluate the color contrast of football jerseys used in the 2024 Euro Cup.

19 Jul 2024 • 12 min read

Tomato Leaf Disease Detection and Diagnosis using Computer Vision

Learn how to build a tomato leaf disease detection and diagnosis system with computer vision.

19 Jul 2024 • 8 min read

People Counting Using Computer Vision

SUMMARY Computer vision-based people counting tracks entry and exit flows in real time, giving organizations accurate occupancy data that manual headcounts cannot match at scale. This tutorial uses a people detection model from Roboflow to build a working counter that draws bounding boxes, overlays a running tally on each

16 Jul 2024 • 8 min read

What is the Open Images Dataset? A Deep Dive.

SUMMARY Open Images V7, released by Google in 2022, contains over nine million annotated images spanning nearly 20,000 categories and supports six annotation types: image-level labels, bounding boxes, segmentation masks, visual relationships, localized narratives, and point-level labels. That breadth makes it one of the few public datasets

11 Jul 2024 • 11 min read

How to Train RT-DETR on a Custom Dataset with Transformers

SUMMARY RT-DETR (Real-Time DEtection TRansformer), developed by Peking University and Baidu, is a transformer-based object detection model that targets competitive speed and accuracy compared to YOLO-family models. This tutorial covers fine-tuning RT-DETR on a custom dataset sourced from Roboflow Universe, using the HuggingFace Transformers

10 Jul 2024 • 16 min read

What is Thresholding in Image Processing? A Guide.

Learn what image thresholding is and the thresholding strategies you can use in computer vision applications.

10 Jul 2024 • 7 min read

The Guide to AI OCR

Learn what AI OCR is and how it is used in computer vision.

10 Jul 2024 • 5 min read

How to Use Florence-2 for Optical Character Recognition

Learn how to use the Florence-2 model for Optical Character Recognition tasks.

10 Jul 2024 • 5 min read

What Is Dense Image Captioning?

Learn what dense image captioning is and how to use the MIT-licensed Florence-2 model to generate dense image captions.

10 Jul 2024 • 5 min read

What Is FPS? A Computer Vision Guide

Learn what FPS is and what FPS considerations you should keep in mind when working on computer vision projects.

9 Jul 2024 • 7 min read

What is 4M? Apple's Massively Multimodal Masked Modeling

SUMMARY Apple's 4M (Massively Multimodal Masked Modeling) framework trains a single unified Transformer encoder-decoder across text, images, 3D geometry, and semantic data by mapping all modalities to discrete tokens and applying a masked modeling objective on a random subset of them. Unlike specialized vision models constrained to

9 Jul 2024 • 5 min read

How to use Florence-2 for Instance Segmentation

SUMMARY Florence-2 is a lightweight MIT-licensed multimodal model from Microsoft that supports instance segmentation through text prompts, without requiring task-specific fine-tuning for many common use cases. This tutorial walks through loading Florence-2 in a Colab notebook, writing a segmentation function using the REFERRING_EXPRESSION_SEGMENTATION

5 Jul 2024 • 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

28 Jun 2024 • 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

27 Jun 2024 • 8 min read

What is F1 Score? A Computer Vision Guide.

Learn what F1 score is, for what it is used, and how to calculate F1 score.

20 Jun 2024 • 5 min read

Florence-2: Vision-language Model

Florence-2 is a lightweight vision-language model open-sourced by Microsoft under the MIT license.

25 Sep 2023 • 6 min read

What is DETR (Detection Transformers)?

In this guide, we discuss what DETR is, how it works, the strengths and disadvantages of DETR, and how DETR performs.