Latest Posts

15 Sep 2021 • 2 min read

Introducing the Roboflow pip package

SUMMARY The Roboflow Python pip package gives developers a native way to interact with their Roboflow workspace directly from Python scripts and Jupyter notebooks. With it, you can access workspace and project metadata, upload images, run inference on trained models, and visualize or save predictions programmatically. Authentication uses your existing

9 Sep 2021 • 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

6 Sep 2021 • 3 min read

How to Implement Object Tracking for Computer Vision

This post is a comprehensive guide on how to implement object tracking with your object detection model to track your custom objects

2 Sep 2021 • 2 min read

Our New REST API

SUMMARY Roboflow's REST API gives developers programmatic read access to workspace data without leaving their code environment. The initial launch covers five routes: API key validation, workspace info, project details, version metadata, and export download links. Combined with the Upload, Annotate, and Inference APIs, these endpoints enable integrations

2 Sep 2021 • 1 min read

Roboflow Changelog: September 2021

Welcome to another installment of the Roboflow Changelog where we summarize the improvements to Roboflow's product suite over the preceding month. You can find the last changelog (for August 2021) here. The biggest highlights from last month were celebrating the release of Roboflow Universe, our community space for

30 Aug 2021 • 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

29 Aug 2021 • 1 min read

Live Coding: Blackjack Basic Strategy

SUMMARY This live coding session builds a computer vision web app that reads playing cards from a webcam and displays real-time blackjack basic strategy recommendations. Starting from boilerplate sample code, the session uses a playing cards object detection model from Roboflow Universe and produces a working app in about

23 Aug 2021 • 3 min read

What is Zero Shot Object Tracking?

We are exciting to announce that you can now track objects frame over frame in video and camera stream using the Roboflow Inference API and the open source zero shot object tracking repository, without having to train a separate classifier for your object track features.

19 Aug 2021 • 6 min read

ImageNet contains naturally occurring NeuralHash collisions

SUMMARY NeuralHash, the perceptual hashing model behind Apple's client-side CSAM detection system, is designed so that two distinct images should never share the same 96-bit hash. A search through ImageNet uncovered two naturally occurring image pairs with identical NeuralHashes, independent of the adversarially crafted collisions researchers

18 Aug 2021 • 3 min read

Mitigating the Collision of Apple's CSAM NeuralHash

SUMMARY Researchers reverse-engineered Apple's NeuralHash model and demonstrated that adversarially crafted images could be made to share a hash with an unrelated target image, potentially flooding Apple's human review queue with false positives. Running the colliding image pairs through OpenAI's CLIP shows that

16 Aug 2021 • 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

11 Aug 2021 • 4 min read

Roboflow Universe: the Computer Vision Community

We’re excited to introduce Roboflow Universe, the new hub for computer vision datasets and pre-trained models.

7 Aug 2021 • 1 min read

Roboflow Changelog: August 2021

Each month we share a recap of our monthly updates and releases in the Roboflow Changelog. The previous Changelog is available here. In July, the product team's major emphasis was on a new product we'll be releasing at our Launch Party on Wednesday, August 11 - hope

7 Aug 2021 • 5 min read

Using Computer Vision to Accelerate Microbiology Research and Combat Antibiotic Resistance

Michael Shamash is a Master’s student in the Maurice Lab at Canada's McGill University. Michael used Roboflow to streamline the production of a model and iOS app for use in microbiology.

6 Aug 2021 • 3 min read

Transformers Take Over Object Detection

It seemed just like a matter of time... and now the Transformers neural networks have landed - Microsoft's DyHead achieves state of the art object detection using a Transformer backbone.

2 Aug 2021 • 7 min read

How to Train YOLOX On a Custom Dataset

The YOLO family continues to grow with the next model: YOLOX. In this post, we will walk through how you can train YOLOX to recognize object detection data for your custom use case.

1 Aug 2021 • 1 min read

What is ONNX?

Learn what ONNX is and when it i sused in computer vision.

30 Jul 2021 • 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.

25 Jul 2021 • 9 min read

Experimenting with CLIP and VQGAN to Create AI Generated Art

SUMMARY Combining OpenAI's CLIP model with VQGAN allows a generative network to be steered toward a text prompt by using CLIP as a scoring signal during image generation, replicating some of the behavior behind DALL-E before that model was publicly released. This post documents a series of

21 Jul 2021 • 2 min read

Announcing On-Prem and Offline Mode for Roboflow Deploy

SUMMARY Organizations handling sensitive data such as private healthcare images, security camera feeds, or classified infrastructure footage need computer vision inference to stay inside their private networks. Roboflow's on-premise inference server runs as a Docker container inside a private cloud or on dedicated hardware, requires only a

19 Jul 2021 • 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

16 Jul 2021 • 4 min read

5 Reasons to not Fully Outsource Labeling

So you're working on building a machine learning model, and you have hit the realization that you will need to annotate a lot of data to build a performant model. In the machine learning meta today, you will be bombarded with services offering to fully outsource your labeling woes.

16 Jul 2021 • 4 min read

Solving the Out of Scope Problem

When we are teaching a machine learning model to recognize items of interest, we often take a laser focus towards gathering a dataset that is representative of the task we want our algorithm to master.

12 Jul 2021 • 11 min read

How AI Protects My Garden from Rabbits

Rabbits were eating all of my vegetables. I decided to take a stand and implement a computer vision enabled system to automatically spook them away from my garden.