Data Asset Catalog From scarce real-world data to the AI intelligence foundation

70,000 medical journal articles with richly illustrated medical literature, clinical case reports and imaging captions, with full sub-licensing rights.

5M real-world structured physical examination data, gender-balanced 1:1 across multiple age strata, for populations with exam fees above 1000 CNY.

Massive authoritative medical knowledge: medical textbooks, clinical guidelines, review articles and drug package inserts.

5M 12-lead ECG waveform images and diagnostic reports including automated analysis reports and physician-reviewed diagnoses.

5M chest X-ray plain-film DICOM images and structured reports; double-read gold standard.

5M head CT plain + enhanced DICOM series and diagnostic reports covering cerebral infarction, hemorrhage and intracranial tumors.

Whole-slide digital images (≥20× scanning magnification) and full pathology diagnosis reports with IHC results and TNM staging.

5M endoscopy videos, key-frame snapshots and diagnostic reports, including polyp location/size and classification.

5M full-diagnostic-pathology image + text datasets covering the complete diagnostic workflow.

5M difficult-disease MDT consultation texts for rare/complex disease AI training — scarce high-difficulty case data.

5M Qupath fine region annotations with six label classes: HCC / ICC / cHCC-ICC / TLS / MVI.

1M liver lesion image annotations for differentiation between HCC and hemangioma.

20M full-cycle inpatient text/PDF documents covering medical-record QA, DRG and DIP.

5M+ medical entities, 20M+ relation triples, ICD-10/11 & SNOMED CT mappings across 7 knowledge bases.

20M+ medical imaging training samples covering CT/MRI/X-ray/US/endoscopy/pathology with classification, detection and segmentation annotations.

3M+ structured disease-specific registry data across 26 key cohorts including diabetes, stroke and oncology; 5+ years long-term follow-up.

328,000 NER and relation extraction annotation records from EHRs, diagnostic reports, physician orders and literature abstracts.

52,800 DICOM images with specialist physician annotations.

Cross-modality medical imaging universal annotation dataset across CT, X-ray, ultrasound and MRI.

1M+ annotated surgical video frames with 24-class anatomy and instrument pixel-level segmentation.

52,800 DICOM images with specialist physician annotations.

1M+ tongue/facial images and 50,000+ specialist TCM annotations.

1M+ DICOM image slices covering 7 major organs and thin-slice CT at 512×512+ resolution.

5M+ China drug data entries across 15+ core dimensions and full ATC 5-level hierarchy, supporting rational-drug-use AI and pharmacovigilance.

1M+ medical device registration and consumable data entries covering registration numbers, management classes, indications and technical specifications.

10M+ medical insurance settlement lists and payment data across 14+ core dimensions, covering formularies, DIP grouping, DRG coding and cost composition.

Cross-industry CAD 2D/3D drawing datasets for drawing understanding, BIM conversion and industrial large model training.

Contract texts across sales, leasing, services and labor types supporting contract review and risk-identification model training.

Cross-industry digital-solution texts and case data supporting industry knowledge bases and solution-generation model training.

Supplier relations, purchase orders, logistics nodes and inventory data for supply-chain optimization and risk warning.

Land expropriation scope, compensation standards, ownership information and approval workflow data for land management and compliance auditing.

Urban operations vital signs, grid events and public service data for city governance models and smart decision support.

Standardized multi-discipline test banks with solutions for smart test generation, error diagnosis and personalized learning.

Historical real exams, mock tests and candidate answer data, suitable for education assessment and intelligent grading.

Science & technology policy, patents, technical standards and industry trend data supporting intelligence analysis and AI4Science applications.

Expert-level high-difficulty multi-disciplinary Q&A pairs for evaluating LLM reasoning capability and knowledge boundaries.

Security model detection boundary cases and false-positive datasets for defensive blind-spot identification and robustness improvement.

C/C++ buffer overflow vulnerability code and triggering input samples supporting binary security and static analysis model training.

Multi-source log, traffic and alert correlation samples supporting SOC intelligent triage and ATT&CK tactic/technique recognition models.

Command injection, deserialization and SSTI attack samples covering multi-language/framework RCE payloads and defense rules.

Reflected, stored and DOM-based XSS payload collections covering encoding bypass and CSP circumvention scenarios for front-end security testing.

Common-law multiple choice, case analysis and precedent reasoning questions for legal-education AI and bar-exam models.

Massive Chinese court judgment corpus covering cause of action, facts, reasoning and judgment outcomes for similar-case recommendation and sentencing prediction.

Structured national/local laws, administrative regulations and judicial interpretations supporting statute retrieval and compliance QA.

Common legal Q&A paired with professional lawyer answers, suitable for legal customer service and LLM fine-tuning.

Judgment summarization and dispute-focus extraction supporting quick case reading and abstract-generation models.

Case facts → applicable law → judgment outcome chain-of-thought data for judicial reasoning LLM.

CN-EN and multilingual legal parallel corpora supporting cross-border legal translation and international compliance.

Multi-view object/scene photo collections with camera poses, intrinsics and sparse reconstruction. Suitable for NeRF/3DGS/Gaussian Splatting reconstruction and world model training.

UAV aerial orthophoto and oblique photography data with POS, RTK coordinates, gimbal angles and large-area scene Meshes — suitable for city-level 3D reconstruction and simulation-environment construction.

Paired data of input photos and corresponding high-quality 3D models with textures & materials for image-to-3D generation, monocular reconstruction and AIGC 3D production.

High-precision 3D scene data for parks, campuses, factories, historic buildings and urban blocks with semantic segmentation, instance labels and interactive objects for digital twins and simulation training.