# Custom Vision Build custom computer vision models from labeled images and use them to tag new images through a REST API. - Findings24: - Site: - Categorias: Implantação de modelos de IA, Reconhecimento de Imagem IA ## O que é Custom Vision? Custom Vision is a computer vision tool for creating models tailored to a specific image-based use case. Users provide examples of labeled images, teach the model the concepts they care about, and evaluate the resulting model before applying it to new images. The workflow combines image labeling, model training, and evaluation. Once a model is ready, simple REST API calls can be used to tag images with the custom model. Microsoft has announced the planned retirement of Azure Custom Vision, with full support for existing customers until September 25, 2028. Existing customers should plan their migration to alternative solutions. ## O que Custom Vision pode fazer? ### Labeled image upload Bring labeled image examples into the service as the starting point for a custom computer vision model. ### In-product image tagging Add tags to unlabeled images within Custom Vision so they can be used as training examples. ### Custom model training Train a model on labeled images to teach it the specific visual concepts required for a use case. ### Model evaluation Evaluate the trained model before using it to process new images. ### REST API image tagging Send images through simple REST API calls and receive tags from the trained custom model. ## Casos de uso ### Recognize application-specific image categories Create a model around visual concepts that are specific to an organization or application, using representative labeled examples. ### Turn an unlabeled image collection into training data Add tags to existing unlabeled images inside the workflow, then use those examples to train a model. ### Add custom image tagging to an application Evaluate a trained model and connect it to an application or service through REST API calls for image tagging. ## Perguntas frequentes ### What does Custom Vision do? It creates custom computer vision models from labeled image examples and uses those models to tag images. ### Do I need labeled images to get started? The workflow supports bringing labeled images or adding tags to unlabeled images before training. ### How are models trained? Users provide labeled images and train the model to recognize the visual concepts relevant to their use case. ### How can a trained model be used? After evaluation, images can be tagged with the model through simple REST API calls. ## Informações rápidas - **Category**: Computer vision - **Core workflow**: Upload or tag images, train, evaluate, and tag images - **Input**: Labeled image examples - **Output**: Image tags from a custom computer vision model - **API**: REST API