DOCSAID
A playground for our developers.
Open Source Projects
Explore our open-source projects and drive AI innovation with us.
Technical Documentation
We share insights to clarify our vision and enhance development.
Research Analysis
We share notes to reveal our thoughts from research papers.
Featured Projects
DocClassifier
Image classification system integrating contrastive learning and PartialFC.
Learn More →
Chameleon
A deep learning toolbox based on PyTorch, integrating Timm and custom training modules.
Learn More →
GmailSummary
Automates email summarization and organization with Gmail API and OpenAI API.
Learn More →DocClassifier
Image classification system integrating contrastive learning and PartialFC.
Learn More →Chameleon
A deep learning toolbox based on PyTorch, integrating Timm and custom training modules.
Learn More →GmailSummary
Automates email summarization and organization with Gmail API and OpenAI API.
Learn More →Recent Updates
👍 Follow our Facebook Page
🤝 Model-Centric Consulting & Services
A small model-focused studio. We turn real needs into maintainable, deployable, and evolvable model modules, working embedded with your team. Frontend/backend are lightweight—only to evaluate, showcase, and integrate models.
🧩 Module Dev & Maintenance
Productize one model module and maintain it long-term with versioned data & benchmarks
🗓️ Consulting
1–2 days/week embedded collaboration: experiment design, data governance, evaluation
🚀 MVP from Zero
Build a minimal, demonstrable product around the model: selection, API, lightweight UI
⚡ Inference Optimization & Deployment
ONNX/TensorRT, quantization, latency budgeting; SDK/REST, batch/stream, private deploy
📚 For full service details, deliverables, and budget ranges, please visit: Full Service Overview
- 🧠 Model-first scope: FE/BE are lightweight for evaluation/showcase/integration only
- 🗓️ Ongoing work often uses a monthly retainer; final quotes follow scoping
- 💵 Billing in TWD (Taiwan).
- 🌍 Cross-timezone/international work is welcome—please schedule in advance
- ❌ We don’t offer LLM self-training (consultation and system evaluation are OK)
- 📜 NDA supported; change logs and rollback strategy provided
Cooperation Form
Please fill in the following information. I will reply within 1 to 2 business days.
Features
Below are demos of two model modules: DocAligner for document alignment and MRZScanner for MRZ recognition.
DocAligner Demo
Upload an image containing a document to detect key points and perform perspective correction.
Test Images
Text Interference
Partial Occlusion
Strong Reflection
Low Light Scene
Highly Skewed
Demo
MRZScanner Demo
Upload an image with MRZ (Machine Readable Zone) to detect and parse the text content.
Test Images
Dim Lighting
Office Desk
Outdoors
Interference
Highly Skewed
Demo
Testimonials
"Thank you for sharing your programming knowledge with such passion and clarity."
— Alexis
"非常感谢分享的文章。"
— cactusgame
"非常实用的功能。"
— frankelinli