Subject recognition in Bead Studio v1.3

U2-Net: Going Deeper with Nested U-Structure for Salient Object Detection.
Xuebin Qin, Zichen Zhang, Chenyang Huang, Masood Dehghan, Osmar R. Zaiane, Martin Jagersand. Pattern Recognition 106 (2020), 107404.
Official repository: https://github.com/xuebinqin/U-2-Net
Model: unmodified lightweight U2-NetP ONNX as distributed by rembg.
Source: https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2netp.onnx
Expected MD5: 8e83ca70e441ab06c318d82300c84806
Upstream U2-Net is Apache-2.0 licensed; see U2NET-LICENSE.txt.
Input/output normalization follows the public rembg U2netpSession implementation (MIT): https://github.com/danielgatis/rembg
Our browser orchestration and mask compositing are project-specific implementations.

ONNX Runtime Web 1.20.1 is Copyright (c) Microsoft Corporation, MIT license.
See ../vendor/ort/LICENSE.txt and ../vendor/ort/ThirdPartyNotices.txt.
The model and runtime are self-hosted public assets. Inference happens on the visitor's device; no image, mask or project is uploaded.