| Pre-trained image analysis | Yes, 250,000+ pre-trained models and 1,000,000+ datasets on Roboflow Universe | Yes, labels, faces, face liveness, text, unsafe content, celebrities, and PPE | Yes, labels, objects, text, faces, logos, landmarks, explicit content, and web entities |
| Custom model training | Yes, RF-DETR, YOLO (v8 / v11 / v26), Detectron2, and more | Custom Labels; trains an AWS-managed model; architecture choice not documented | Separate service; AutoML in Gemini Enterprise Agent Platform (formerly Vertex AI); legacy AutoML Vision shut down July 31, 2024 |
| Custom vision tasks | Object detection, instance and semantic segmentation, classification, keypoints, OCR, depth estimation, and multimodal tasks | Classification and detection only; AWS says Custom Labels is not designed for faces, text, or unsafe content | Classification and object detection with AutoML image training |
| Labeling your images | AI-drafted; Auto Label labels the classes you describe in a text prompt; Smart Polygon powered by Meta AI's Segment Anything; Label Assist pre-annotates with your own trained models | By hand; image-level labels and bounding boxes drawn in the console, or import a SageMaker Ground Truth manifest | Manual; labels are added in the Google Cloud console, and the Data Labeling Service shut down October 3, 2024 |
| Keeping models accurate | Built in; active learning and operator feedback send flagged production images back into your dataset, and Vision Events logs every prediction and image | Manual; add and label new images, then train a new model version, or install the separate Model Feedback sample solution from GitHub | Manual; Cloud Vision API models are fixed; custom models are relabeled and retrained in a separate service |
| Export and own model weights | Yes | Not documented; models are called through the Custom Labels API | AutoML Edge only; not documented for Cloud Vision API models |
| Edge and offline deployment | Yes, NVIDIA Jetson, GPU servers, Raspberry Pi, x86, or Roboflow AI1, online or fully offline | AWS cloud only; Custom Labels models run on inference units in an AWS Region; edge or offline use not documented | Partial; exported AutoML Edge models only; Cloud Vision API is a cloud REST and RPC API |
| Video and cameras | Any RTSP, USB, GigE, or industrial camera | Stored video, analyzed from Amazon S3; Streaming Video closed to new customers as of April 30, 2026 | Images and files; video analysis is a separate product, the Video Intelligence API |
| Running a trained model | Serverless hosted API that scales with traffic, or self-hosted on your own hardware, with no model to start or stop | Start and stop each model version; inference units bill per hour while running, even when idle | Cloud Vision API runs on Google's endpoints; custom AutoML models are served through Google Cloud |
| Self-hosted inference | Yes, open-source inference server for custom and foundation models | No; Custom Labels models are served as AWS-managed endpoints billed in inference units | No; the Cloud Vision API is not self-hostable |
| Uptime SLA | Yes, on Enterprise, with a public status page | Yes, 99.9% monthly uptime per AWS Region, with service credits | Yes, 99.9% monthly uptime for Cloud Vision, with financial credits |
| Compliance | SOC 2 Type II, HIPAA with BAAs, PCI DSS | Assessed under AWS programs including SOC, PCI, FedRAMP, and HIPAA | In scope for ISO 27001 / 27017 / 27018, SOC 1 / 2 / 3, and PCI DSS; HIPAA-covered under Google's BAA |
| SDKs and APIs | Yes, Python SDK, REST API, MCP server, public docs; open-source inference server and supervision library | Yes, AWS SDKs (boto3 for Python, plus Java, .NET, Go, JavaScript, Ruby, PHP) and the AWS CLI | Yes, Google Cloud client libraries for Python, Node.js, Go, Java, Ruby, PHP, and C#, plus REST and gRPC APIs |
| Pricing | Published plans; every plan, including Free, includes self-hosted inference, model weights download, and a private workspace | Image APIs from $0.001 per image for the first 1 million images a month; Custom Labels $1 per training hour and $4 per inference hour | Billed per image per feature; first 1,000 units per feature free each month, then $1.50 per 1,000 for most features |