Face Training and Analysis Powered by DeepVA Motivation [VN UG]
Train a custom model and analyze your videos for known faces
Provides the ability to train a model on faces stored in VidiCore and then analyze video content for detected faces.
Using image assets containing faces as training material you can now create a custom face detection model provided by DeepVA, and based on this model you can run an analysis of your video content to detect faces known to the model and automatically populate the assets with metadata from the DeepVA face detection model. The service will also return keyframes of faces still unknown to the trained model, which could be used to further enhance the training model. Media files to be analyzed must be under management by VidiCore and have thumbnails generated for them.
For all customers looking to go beyond what can be achieved by pre-trained cognitive services and wants to build their own model of faces that might not be known to an international public. This could for instance be a regional news channel building a face detection model of regional politicans and celebrities, in order to easily find and fetch video material containing them when creating a news segment.
Seamless integration into any VidiCore system, anywhere.
Prerequisites
VidiCore version 21.3 or later
Callback-bucket or directory on cloud storage with VCS access configured. Instructions provided in knowlegebase.
Training assets residing on supported cloud storage (AWS S3)
Video assets residing on supported cloud storage (AWS S3)
Service Activation
Activate the service by posting the ResourceDocument to your VidiCore Server on /API/resource/vidinet.
Connection Status
Connection status can be seen in VidiCore Server from /API/resource/vidinet and in the Vidinet dashboard -> Service details. You can connect multiple VidiCore Server instances to one VidiNet Service, using the same or different access key pairs. (Using unique key pairs for each VidiCore Server is recommended).
Service Usage
For detailed instructions on how to run face training and detection jobs from the VidiCore Server API see the Face Training Theme and the VidiCore API documentation.
Job Status and Billing
VidiCore Server will provide Vidinet job status and progress in /API/job and in the Vidinet dashboard -> Service details.
Successful jobs will be charged to your account and listed in the Vidinet dashboard -> Billing. Failed jobs will not be listed nor billed.
Supported formats
For list of supported formats, see DeepVA documentation.
Support
Visit support.vidispine.com and we will assist you.