Marcelle: Composing Interactive Machine Learning Workflows and Interfaces
summary
Marcelle is an open-source JavaScript toolkit (MIT) that composes interactive ML workflows from modular components embedding data, models and GUIs; it implements reactive pipelines for event streams, supports real-time synchronized data stores via FeathersJS, and offers dashboard and wizard layouts for custom interfaces. Partial TensorFlow/Python interoperability lets engineers log training, drag‑and‑drop images for instant or batch prediction, and share models with clinicians; two implemented case studies include novice interaction studies and a skin‑lesion diagnosis workflow showing end‑to‑end collaborative prototyping.
AI-written summary · about
chapters (5)
- 0:00Introduction to Marcel
- 1:28Interactive ML challenges
- 2:36Architecture design principles
- 6:10Skin cancer diagnosis study
- 7:39Conclusion and future work
related talks
46mArtificial Intelligence with JavaScript
10mInteractML: Making machine learning accessible for creative practitioners working with movement ...
43mMachine Learning Engineering principles with Python and MLFlow
22mProductive Machine Learning and Deep Learning Projects
34mOrchestrating Generative AI Workflows to Deliver Business Value
36mPractical Machine Learning for Front End Developers
33mPrototype ML with Visual Blocks
25mHands-on Multimodal AI Development with Pixeltable