# Defensive Publication: Rūpestėlis ID — Smartphone-Captured Pet Biometric Health Card with End-to-End CNN Embedding and Cryptographically Signed Registry **Publication ID:** RID-PA-2026-001 **Author:** Rūpestėlis Holding (architect: Dr. Tomas Margelis, DVM) **Published:** 2026-04-25T14:21:56.948926+00:00 **Cryptographic priority:** Ed25519 / Sigillum **Content hash (SHA-256):** `26a8ca56882e2d8e49dad37feaeca0222de062821c28bf108d9c817848a049b7` **Signature:** Sigillum Ed25519 — see /sigillum/public-key for verification **Persistent URL:** https://rupestelis.com/prior-art/RID-PA-2026-001 --- ## Purpose of this Disclosure This document is published to establish **prior art** under EPC Art. 54(2) and 35 USC §102(a)(1). By making the technical content below publicly available with a cryptographically verifiable timestamp, we ensure that: 1. No third party can validly patent the disclosed methods after this date. 2. Rūpestėlis Holding can demonstrate priority in any future dispute. 3. The technical community has free access to use, study, and build upon these methods. This is **defensive publishing** — we are explicitly choosing NOT to patent this technology, but to make it provably ours-first via public disclosure. --- ## Technical Description ## System Overview Rūpestėlis ID is a pet health card registry that uses smartphone-captured nose-print biometrics as the primary identifier, combined with cryptographically signed health records, to provide a portable, vet-readable, theft-resistant pet identity that complements (not replaces) microchip identification. ## Pipeline Architecture ### 1. Capture Layer (smartphone-only) - Standard smartphone camera (no specialized hardware) - On-screen UX overlay (alignment circle as soft cue, not physical stabilizer) - Image normalization in software (white balance, exposure) - Optional liveness check via short video clip (movement = real animal) ### 2. Feature Extraction (end-to-end CNN, no intermediate keypoints) - Backbone: EfficientNet-B3 OR DINOv2 ViT-L (no architectural modification) - Trained with triplet loss + ArcFace margin - Output: 512-dimensional dense embedding vector - NO Gabor transforms. NO frequency-transform binarization. - NO explicit keypoint detection. NO geometric region carving. - NO alignment-then-feature pipeline. The CNN handles alignment via learned spatial attention internally. ### 3. Matching (vector-space similarity) - Cosine similarity between query embedding and registry embeddings - Faiss-style approximate nearest neighbor index for scale - NO block-wise shift matching. NO frequency-domain operations. ### 4. Registry & Signing (cryptographic priority) - Each enrollment record: pet metadata + nose embedding + health records - Ed25519 signature (Sigillum scheme) over canonicalized JSON - Sigillum chain: each event references prior_event_id - PostgreSQL storage with foreign key integrity ### 5. Health Card Layer (the real value) - Health records (vaccinations, surgeries, allergies, medications) attached to the biometric ID - Vet Quick Scan: scan nose → see emergency-relevant medical history in 3 sec - Cross-clinic portability without paperwork - Owner-controlled access ## Why This Pipeline Avoids Existing Patent Claims This architecture was designed deliberately to operate outside the claim scope of: - iSciLab EP3029603B1 (Gabor + block-shift + hardware stabilizer) - Petnow EP4447007B1 (keypoint + circle/ellipse + alignment-first) By using end-to-end CNN embedding without intermediate keypoint geometry, without frequency-domain operations, and without physical stabilization hardware, this pipeline operates on a completely different technical basis. ## Implementation Status - Schema: `rupestelis_id.pets`, `nose_prints`, `health_records`, `sigillum_events` (PostgreSQL 16, deployed 2026-04-25) - Backend: FastAPI, 10 REST endpoints, 1285 LOC - Cryptographic layer: Ed25519 via Python `cryptography` library, key shared with AGDO Sigillum infrastructure - Frontend: Mobile-first HTML, Pet Health Card positioning - License: Public-good components under CC-BY-SA, commercial integration via rupestelis.com/id ## Public Code Reference The full implementation is documented at: - https://github.com/rupestelis/rupestelis-id (planned 2026-Q3 OSS release) - Pre-release access: contact tomas@rupestelis.com --- ## Novel Aspects Claimed as Prior Art 1. Combination of smartphone-only nose biometric capture, end-to-end CNN embedding (no intermediate keypoints), cosine-similarity matching, and Ed25519-signed registry into a single PET HEALTH CARD product (not anti-theft, not lost-pet) — novel positioning. 2. Vet Quick Scan: public-key-readable emergency medical info exposed via nose-print scan without owner authentication, while full medical history requires authentication. Two-tier disclosure model is novel for animal biometric registries. 3. Sigillum chain audit trail linking each pet record event (registration, health update, ownership transfer) to the prior event via cryptographic content hash. Makes the registry tamper-evident at the individual-record level. 4. Country-tier pricing tied to ISO country code at registration, automatically applied via PostgreSQL seed table — operationalizes purchasing-power parity for an EU pet registry. 5. 'Free Heart' enrollment slot: every owner gets one pet registered free, programmatically enforced via single-claim flag in user table. Combines emotional hook with conversion mechanism. 6. Hybrid public/private brand: AI-augmented persona (Dr. Rupi) for consumer interaction with a real human architect (Dr. Tomas Margelis) for legal/regulatory roles, both transparently disclosed in the footer of every certificate. --- ## Related Public Work (cited prior art) - Choi et al., 'Dog Identification Based on Noseprint Recognition Using Deep Learning', IEEE Access 2021 - Kumar, 'Survey of Animal Biometric Identification', Pattern Recognition 2024 - DINOv2 (Meta AI, 2023): https://github.com/facebookresearch/dinov2 - EfficientNet (Tan & Le, 2019): arXiv:1905.11946 - ArcFace (Deng et al., 2019): arXiv:1801.07698 - Faiss (Facebook AI, 2017): https://github.com/facebookresearch/faiss --- ## Search Keywords (for discoverability) pet biometric identification, dog nose print recognition, smartphone pet ID, veterinary health card, Ed25519 pet registry, noseprint CNN, Rūpestėlis ID, Dr. Rupi, pet biometric EU, pet health card mobile --- ## Verification This document is signed with the Rūpestėlis Sigillum Ed25519 keypair. Anyone may verify the signature against the public key at: https://rupestelis.com/sigillum/public-key The signature attests that the content hash above corresponds to this exact document at this exact timestamp. --- ## License This disclosure is released into the **public domain** (CC0 1.0 Universal) to maximize its prior-art weight. Anyone may use, modify, or build upon this technology without restriction. Attribution is appreciated but not required. --- *Defensive publication generated by IP-GUARDIAN-001* *Rūpestėlis Holding — "1 human + AI team = future standard"*