SvelteML 0.0.5

Overview 🤖 📖

The purpose of SvelteML is to offer simple Components that can make ML more accessible. It leverages TensorflowJS to offer Svelte apps ML features out-of-the-box. It relies heavily on Svelte's reactivity feature and event hooks can be used to extend out the ML flow. e.g. on:poses in the Pose Estimator will give you the raw poses directfrom TensorflowJS.

Quick Start

npm install svelteml --save

What's inside?

Classification / Segmentation

• [x] Image Classification
• [x] Body Segmentation
• [x] Basic Multi-Pose Estimation
• [x] Object Detection

Text-based inference

• [x] Sentence Encoding
• [x] Text Toxicity
• [x] Question and Answers

Image Effects

• [x] Blur Body Parts
• [x] Bokeh Effect
• [x] Face Mesh
• [x] Hand Pose Detection

Experiments

• [ ] Switching to Lerna for multiple repos so the lib can expand in the different areas. Also helpful for tfjs3 when it will have code-splitting 😃
  • [ ] @svelteml/ui
  • [ ] @svelteml/classification
  • [ ] @svelteml/segmentation
  • [ ] @svelteml/automl
  • [ ] @svelteml/text
  • [ ] @svelteml/audio
• [ ] Unlock slots with Facial recognition, maybe use faceapi.js
• [ ] Demo site for more details
• [ ] Audio and speech recognition features
• [ ] Additional models using the lower level tfjs apis
• [ ] .... and a few other top secret ideas 🤭

Note

All Components try to be reactive so although it feels very declarative, it is also reacting to your input. Add an issue in Github if you need a specific behaviour or if there is a bug or would like to recommend something. You know the drill.