Custom Vision is a tool or service that can be used for training models for image classification and for detecting objects in images. With Custom Vision, we are able to upload our own images and tag them so that they can be trained for image classification. If we train a model for object detection, we can also tag specific areas of an image. Because models are already pre-trained for basic image recognition, we don't need a large amount of data to get a great result. The recommendation is to have at least 30 images per tag.
When we have trained a model, we can use it with an application programming interface (API) that is part of the Custom Vision service. We can also export models for Core ML (iOS), TensorFlow (Android), the Open Neural Network Exchange (ONNX) (Windows), and to a Dockerfile (Azure IoT Edge, Azure Functions, and Azure ML). These models can be used to carry out classification or object detection without...
