Roboflow Blog

24 Jun 2022 • 2 min read

What is cuDNN?

SUMMARY cuDNN (CUDA Deep Neural Network library) is NVIDIA's GPU-accelerated library of low-level primitives, such as convolution and pooling, that frameworks like PyTorch and TensorFlow call internally to run deep learning workloads on NVIDIA GPUs. It sits on top of CUDA, which handles general-purpose GPU

22 Jun 2022 • 3 min read

Launch: Label Only Role-based Access Control

SUMMARY Roboflow Annotate now supports role-based access control with a dedicated Labeler role, letting teams scale annotation work outside their organization without exposing project data. Labelers see only the images assigned to them in a Job, while Reviewers approve or reject submitted annotations and return failed batches for edits.

15 Jun 2022 • 9 min read

Train and Deploy YOLOS Transformer On a Custom Dataset

In this post, we showcase training and deploying YOLOS end to end, from labeling your data, to training your model, to deploying your model on AWS for inference.

8 Jun 2022 • 3 min read

Roboflow Now Available in the Microsoft Azure Marketplace

SUMMARY Roboflow is now listed in the Microsoft Azure Marketplace, giving Azure customers a direct path to purchase and integrate Roboflow's computer vision platform alongside their existing cloud infrastructure. Teams already storing images and video in Azure Blob Storage can connect that data to Roboflow's annotation,

6 Jun 2022 • 8 min read

Food Manufacturing Defect Detection with Lucky Charms

SUMMARY This tutorial builds a computer vision system to detect and classify the eight types of Lucky Charms marshmallows (marbits) as a practical stand-in for food manufacturing quality control. Dataset labeling and preprocessing happen in Roboflow, and the post covers two training paths: Roboflow Train for a one-click

6 Jun 2022 • 6 min read

Autonomous Vehicle Object Detection

SUMMARY Ampera Racing, a student engineering team from UFSC in Brazil, built a low-cost autonomous vehicle pipeline using computer vision as the core perception layer, with RF-DETR handling cone detection for the Formula Student Driverless competition. The architecture covers five subsystems: environment perception, mapping, motion planning, lateral control,

19 May 2022 • 2 min read

Launch: Instance Segmentation Project Training and Inference

SUMMARY Roboflow now supports instance segmentation project training and inference natively on the platform, letting teams train directly from a pre-trained COCO checkpoint without writing training code. Instance segmentation produces per-object outlines rather than just bounding boxes, making it the right choice when downstream tasks require measuring object

19 May 2022 • 4 min read

Launch: Roboflow OAK pip package

roboflowoak pip package lets you natively interface with your Roboflow workspace through Python scripts and deploy to an OAK device.

17 May 2022 • 4 min read

Boxing Punch Detection Using Computer Vision

One of the best parts about joining Roboflow is doing a computer vision project in your first 2 weeks. As someone who loves to workout, I wanted to focus my project on a fitness related use case. I often mix boxing workouts into my routine, and thought it would be

16 May 2022 • 9 min read

Recipe for a Company On-site

Creating a magnetic environment for exceptional talent is one of Roboflow’s top priorities. We want our remote-first culture to be the best option for our team, not a necessity because of the pandemic or anything else. We choose to be remote. It works better for us than being

11 May 2022 • 3 min read

How to Train a YOLOv5 Oriented Bounding Box Model

Learn how to train a YOLOv5 Oriented Bounding Box (OBB) object detection model.

11 May 2022 • 5 min read

Roboflow’s Python Pip Package For Computer Vision

Regardless of whether your project is a new product line, a new industrial production system, a research project, or a personal one to help you learn what computer vision is all about, you'll want to add "pip install roboflow" to your code - and here's why.

6 May 2022 • 2 min read

Roboflow Changelog: April 2022

April showers brings May flowers, and here's a bouquet of flower datasets from Roboflow Universe you can use today!

5 May 2022 • 3 min read

Cinco de Mayo, Beer and Taco Dataset with Multi-Label Classification

SUMMARY Multi-label classification lets a single model predict multiple classes per image simultaneously, making it a natural fit for scenes where counting or locating objects precisely is less important than knowing what is present. This post builds a taco and beer classifier by merging several Roboflow Universe datasets (including

3 May 2022 • 3 min read

Machine vision applications with wearable devices

SUMMARY XR Tech, a UK industrial extended reality provider, built a wearable camera safety system by training computer vision models on Roboflow to detect PPE compliance and physical hazards such as loose rubble and heat exhausts on live body-cam streams. The team followed a structured data hygiene process: starting

20 Apr 2022 • 1 min read

Art Recognition with a Computer Vision Model

Learn how to use computer vision to classify art styles.

13 Apr 2022 • 6 min read

How to Train Detectron2 for Custom Instance Segmentation

A walk through on how to train Detectron2 to segment your custom objects from any image by providing our model with example training data.

4 Apr 2022 • 2 min read

Roboflow Changelog: March 2022

Spring is finally here and there is computer vision in the air. Each month, we share a recap of product and company updates; if you missed last month's edition of the Roboflow Changelog, you can find it here. Roboflow Organize and Annotate * Launched Polygonal Annotation tool * Launched Multi-

3 Apr 2022 • 4 min read

Launch: End to End Multi-Label Classification

SUMMARY Multi-label classification lets a model predict several independent attributes for a single image simultaneously, which is fundamentally different from single-label classification and requires far less training data than concatenating class combinations into discrete buckets. Roboflow now supports end-to-end multi-label classification projects, including dataset creation,

2 Apr 2022 • 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.

1 Apr 2022 • 2 min read

Computer Vision Saves 4.95 Billion Internet Users from Rickrolling

Roboflow releases new RICK model to end Rickrolling Newton-le-Willows, United Kingdom, 4/1/2022: — Roboflow, a software business that builds tools for developers to use computer vision, is releasing Real-time Intrusion Checker Kernel (RICK). The model is a breakthrough in machine learning model architecture with its first

24 Mar 2022 • 3 min read

How many images do you need to train a model?

SUMMARY There is no fixed minimum number of images required to train a computer vision model: 100 images can work for a focused, well-defined task, while a complex or high-accuracy use case may need thousands. The more useful framing is to build an MLOps pipeline from the start,

24 Mar 2022 • 5 min read

Improving Computer Vision Datasets and Models

Common tips and tricks to improve your computer or machine vision dataset, including using Roboflow Train and the Dataset Health Check tools.

16 Mar 2022 • 1 min read

Convert Annotation Formats EASY (26 available Formats 😱)

SUMMARY Computer vision datasets come in dozens of annotation formats, and switching between them manually is tedious. Roboflow handles this in two steps: upload your dataset in whatever format you have, and export it in any of 26 supported formats. The platform auto-detects the incoming format, so no manual