mayAI? Exploring Graphical Tools for Model Visualization & Enh.
summary
A hackathon-built tool called mayAI parses neural nets in NNEF text form and renders node-and-dataflow graphs so convolutional, pooling and concat layers are visible, letting teams spot disconnected nodes and bottlenecks. A demo uses an image‑prediction NN and the ccape library to draw graphs; testing shows the approach helps trace failures that are opaque in raw text. The authors cite model repository growth from ~400,000 to ~650,000 models as the scaling problem.
AI-written summary · about
chapters (14)
- 0:00Introduction and team
- 1:08What is mayAI project
- 2:37Why model visualization matters
- 4:22AI model complexity challenges
- 5:51Model formats and neural networks
- 6:54Typical AI model training process
- 8:18How visualization helps development
- 9:22Popular AI visualization tools
- 10:06Building our own tool
- 11:30Commercial tool example
- 12:55Real-world use case benefits
- 13:58Who benefits from visualization
- 15:03Market research findings
- 16:28Q&A on visualization frameworks
related talks
01h 00mData, visualization, and designing AI
01h 26mThinking in Graphs: The Cognitive Foundations of Agentic AI
41mModels Schm-odels: Why you Should Care about Data-Centric AI
01h 03mIt's All About AI
44mKeynote: Tracking the Pulse of Generative AI
48mKeynote – Sameer Singh
44mKeynote: Tracking the Pulse of Generative AI
20mAI and Machine Learning for developers