Google Vision OCR vs Segment Anything Model 2 (SAM 2)

Google Vision OCR, released as part of the Cloud Vision API’s general availability in February 2016, is a proprietary Google Cloud service for extracting text from images and documents. It supports common formats like JPEG, PNG, GIF, TIFF, and PDF, and provides two main modes: TEXT_DETECTION for short snippets and scene text, and DOCUMENT_TEXT_DETECTION for dense documents, which returns structured layout information with bounding boxes.

While not an LLM (so it has no token context window or parameter count), the service performs OCR across printed text and some handwriting. It outputs detected text along with positional metadata, making it useful for digitizing scanned files, receipts, forms, and signs. However, complex layouts like tables often require downstream processing. Accessible via REST and RPC APIs, with client libraries in major languages, Google Vision OCR is widely used for document processing pipelines, archival, and accessibility applications.

SAM 2 is a real-time image and video segmentation model developed by Meta AI, released in July 2024 under the Apache 2.0 license. It extends the original Segment Anything Model to support video inputs by introducing a streaming memory architecture that maintains object state across frames, enabling consistent segmentation of objects through occlusion, motion, and scene changes. For image inputs, SAM 2 operates similarly to its predecessor with improved mask quality and speed.

SAM 2 accepts point, box, and mask prompts and produces object masks interactively or in a fully automated mode. Its memory architecture enables video segmentation at real-time speeds. SAM 2 is used in annotation pipelines, video analysis, robotic perception, and any application requiring high-quality promptable segmentation across both images and video.