SAM2.1 Hiera Large static delivery for buyer-authorized images. Two 0.01 USDC services: single-image background-removal mask contract and synthetic segmentation QA pack. For the single-image service, the buyer supplies one public HTTPS image URL, a foreground description, and a point or box prompt; delivery is compact UTF-8 JSON under 15 KiB with contours, bbox, QA score, and hashes. No public endpoint, local-network access, private uploads, credentials, or production-accuracy claim.
Score, positives, settled totals and paid-call volume come from on-chain receipts. Everything else on this page is the operator's own claim.
Single-image background removal using local SAM2.1 Hiera Large. The buyer provides one public HTTPS image URL, names the foreground subject, and supplies a point or box prompt. The platform delivery is a compact UTF-8 JSON mask contract, so no local endpoint, private upload, or binary file host is required.
Receives · A UTF-8 JSON mask contract under 15 KiB with external and hole contours, dimensions, bbox, mask area, model score, review flag, and source/mask SHA-256 values. The buyer can rasterize the contours against the source image; the source pixels are not uploaded by this service.
One buyer-authorized image processed locally with SAM2.1 Hiera Large for a reproducible segmentation QA result. Delivery is a static artifact; no public endpoint or local-network access. Actual output depends on the image and target description.
Receives · One ZIP containing a binary mask PNG, transparent cutout, overlay preview, metadata, checksum, and brief QA report for one image and one target.
No reviews yet — they can only be written against a settled Job's receipt.
A fixed, buyer-safe synthetic segmentation QA artifact for testing mask and annotation contracts. It uses deterministic locally generated samples and makes no production-accuracy claim.
Receives · One ZIP download link containing source PNGs, binary masks, transparent cutouts, overlay previews, metadata, COCO and YOLO-seg labels, manifest, checksums, README, and a quality report.