Asynchronous processing
Results are returned as asynchronous jobs that you poll for completion. Submit an image for analysis, receive a job ID, and check the job status until it finishes. Then retrieve the extracted features and metadata.Use cases
Vision is useful for inspecting equipment in photos, extracting text from signage or labels, detecting defects in imagery, and automating visual verification tasks. The API supports industrial imaging scenarios common in asset-intensive industries.Ensure images are in supported formats and meet size requirements before submission.
Extract
Extract starts a job that reads text, asset tags, or industrial objects from images. A project can run up to 20 extract jobs at the same time. Each file must be.jpeg, .jpg, or .png, and its mimeType must be image/jpeg or image/png.
Completed batches are returned while other batches are still running. The job stays Running until every batch finishes. A failed item is listed in failedItems and does not drop results from completed batches. The job is Completed when at least one batch succeeds, and Failed when every batch fails.
Segment
Segment starts a job from prompt points, a bounding box, or both. More points on the object improve the result. A point label of1 marks foreground and 0 marks background. File requirements match extract. This operation is available on the beta and alpha APIs.
Key capabilities
- Detect objects in images
- Extract text (OCR) from images
- Identify visual elements for classification
- Segment images (for example region or object segmentation via
/context/vision/segmentand related job endpoints) as asynchronous jobs you poll - Poll job status for async results