-
Book Overview & Buying
-
Table Of Contents
-
Feedback & Rating

Computer Vision on AWS
By :

Once a CV model has been deployed, you should regularly evaluate its performance to establish when retraining is required and perform deployment testing before rolling out a new version. This ensures that models are delivering reliable results and that they are meeting your established business outcomes. Deployment testing helps you detect changes in a model’s accuracy and identify any errors in a model’s implementation before it is deployed to production.
Shadow testing is a testing technique for evaluating the performance of a model before it’s rolled out to production. A new (shadow) ML model is tested in a production environment without impacting actual user traffic. The shadow model’s predictions are not used in the production application; instead, the shadow model runs alongside the existing production model. A copy of the inference requests is routed to the shadow model and its predictions are...