Here are a few things that any organization deploying CV models into production should be doing to ensure that those models continue to perform as expected.
Automatic out-of-distribution detection lets you identify where your model is likely making mistakes.
Explore the results of your vision models (classification and object detection) with an interactive interface that makes it easy to identify issues.
Visualize important image regions that are impactful for model predictions.
Monitor CV model pipelines for data anomalies using built-in out-of-distribution detection and track the accuracy of bounding box models
Detect biases in your CV models by evaluating image classification outputs using an interactive interface and locating where your models misclassify and perpetuate biases
Visualize which regions of an image are impactful for an image classification model’s decision or how your object detection models are performing on pipeline images
Our team has many models in production. With Arthur, checking their output and status at a glance is easy.
– Chris Poirier, VP of Data at Truebill