SenseCraft Model Assistant releases v2.0.0rc3

EdgeLab got a optimized visualization tool for inference, model training and export command-line interface has refactored, also GitHub Actions CI integration is supported now. Please stay turned for upcoming Web tools and C++ SDK for model deployment on edge computing devices.

Highlights

• EdgeLab Model Zoo, which provides some pre-trained models for different kinds of tasks with detailed benchmark information.
• Optimized inference and visualization tool for easy testing and evaluation your trained models on PCs.
• Anomaly detection for gyroscope sensors (Experimental).

New features

• Optimized and refactored the command-line interfaces for model training, export and inference.
• Added CI (GitHub Actions), it is welcome to submit PRs to us.
• Cleaned up redundant code, lots of issues got fixed.

Coming soon

• EdgeLab Web Tools will provide a quick and interactive interface for deploying models to MCUs using browser.
• EdgeLab C++ SDK will deliver a unified and easy-to-use API for rapid deployment of models to edge computing devices (MCUs, SBCs).
• EdgeLab Model Zoo will add more pre-trained models continuously and EdgeLab documentation will be revised accordingly.

Changelog

• Changes between previous release and this release: v2.0.0rc2 - v2.0.0rc3.
• Changes since this release: v2.0.0rc3 - dev.

EdgeLab releases v2.0.0rc2

EdgeLab got a optimized visualization tool for inference, model training and export command-line interface has refactored, also GitHub Actions CI integration is supported now. Please stay turned for upcoming Web tools and C++ SDK for model deployment on edge computing devices.

Highlights

• EdgeLab Model Zoo, which provides some pre-trained models for different kinds of tasks with detailed benchmark information.
• Optimized inference and visualization tool for easy testing and evaluation your trained models on PCs.
• Anomaly detection for gyroscope sensors (Experimental).

New features

• Optimized and refactored the command-line interfaces for model training, export and inference.
• Added CI (GitHub Actions), it is welcome to submit PRs to us.
• Cleaned up redundant code, lots of issues got fixed.

Coming soon

• EdgeLab Web Tools will provide a quick and interactive interface for deploying models to MCUs using browser.
• EdgeLab C++ SDK will deliver a unified and easy-to-use API for rapid deployment of models to edge computing devices (MCUs, SBCs).
• EdgeLab Model Zoo will add more pre-trained models continuously and EdgeLab documentation will be revised accordingly.

Changelog

• Changes between previous release and this release: v2.0.0rc1 - v2.0.0rc2.
• Changes since this release: v2.0.0rc2 - dev.

EdgeLab releases v2.0.0rc1

We upgraded the framework to openMMlab 2.0, and updated the content of the documentation and the framework

Highlights

• Support openMMLab2.0
• Documentation significantly optimized

New Features

• Tflite quantization and export using tinynn
• Documentation framework using vitepress

model zoo

Model weights in different formats，Including pt, onnx, tflite, ncnn。