TrainingData.io® TrainingData.io provides AI-Assisted Image & Video Training Data Labeling at scale. TrainingData.io application is a AI training-data management platform. It empowers data science teams to control quality of AI-training data. It also empowers AI Teams with data collaboration.
In order to train machines to make decisions on behalf of humans, they must learn to make those decisions. In order to learn about the world around us, machines take input in form of images and text that is labeled for representative features. A label in an image is clearly marked bounding region that represents a real world entity or object. Managing data labeling at very large scale is time consuming and requires special software. TrainingData.io application is that software.
To use the TrainingData.io container, you need to SignIn. After you signin, you can manage AI-Training datasets, Labeling instructions, Labeling Jobs from the dashboard. TrainingData.io offers free version of the product to let you try all the power-packed features.
On public cloud instances like Amazon EC2, you can install trainingdataio/tdviewer:v1.0-ngc.
Manage On-Premises TrainingData Labeling
Pull docker image:
docker pull nvcr.io/partners/trainingdataio/tdviewer:v1.0-ngc
Create a directory on your disk to store TD.io database. For example "/home/user/db"
mkdir -p /home/user/db
(Optional) Create a directory to place images and videos (dataset assets). For example: "/home/user/images"
mkdir -p /home/user/images
Run Docker image providing mount point for database and mount point for images folder.
docker run --mount src=/home/user/db/,target=/home/user/trainingdataio/tdviewer/db,type=bind --mount src=/home/user/images,target=/home/user/trainingdataio/tdviewer/images,type=bind -p 8090:8090 -p 9090:9090 nvcr.io/partners/trainingdataio/tdviewer:v1.0-ngc
where
-p 8090:8090
and -p 9090:9090
expose ports 5901 for web-browser connection and 9090 for
REST API connection
--mount
means that local host source disk location will be mounted inside the container at the target location. (for both database and image assets)
How to create on-premises datasets?
How to create labeling instructions?
How to distribute labeling jobs among annotators and reviewers?
Supported export formats for annotated data?
Email: support@trainingdata.io