Machine Learning Tools
NuEdgeWise is a machine learning model training tool.
NuEdgeWise is a graphical user interface (GUI) tool specifically designed for embedded AI development. Users can easily train models through its intuitive interface without requiring deep AI expertise. The tool supports a wide range of AI applications, including Keyword Spotting, Gesture Recognition, Image Classification, Object Detection, Anomaly Detection, and Visual Wake Words. With NuEdgeWise, even beginners can quickly build lightweight and high-performance AI models optimized for MCU platforms, helping developers accelerate embedded AI product development and reduce the barrier to entry.
Use Cases Table
| Tool | Use Case | Model | M55M1 | M467 | MA35D1 | Description |
|---|---|---|---|---|---|---|
| ML_KWS | Keyword Spotting | DNN/DS-CNN | ✔️ | ✔️ | ✔️ | Keyword Spotting uses the intuitive NuEdgeWise graphical user interface (GUI) to help beginners quickly train speech recognition models for identifying specific keywords, such as "turn on" or "turn off." This allows users to easily control devices through voice commands. The system can recognize the configured vocabulary within one second. |
| ML_G-Sensor | Gesture Recognition Magic Wand | CNN | ✔️ | ✔️ | ✔️ | This tool uses the NuEdgeWise GUI interface to train data from a 3-axis accelerometer in order to recognize specific gesture movements, such as waving, rotating, or shaking. It is well-suited for applications like smart home control, wearable devices, or game controllers. |
| ML_Image_Classification | Image Classification | MobileNet/EfficientNet/FD-MobileNet/ShuffleNet | ✔️ | ✔️ | ✔️ | Image Classification uses the GUI tool provided by NuEdgeWise to train image classification models (such as MobileNet) through simple transfer learning. It enables the recognition of different image categories, such as distinguishing between cats, dogs, or other specific objects. Even beginners can quickly get started and train their own image classification models. |
| ML_Object_Detection | Object Detection | SSD_MobileNet_fpnlite v2/v3 | ✔️ | ❌ | ✔️ | The SSD MobileNet fpnlite model for object detection can be quickly trained and deployed using NuEdgeWise's visual interface. It accurately identifies and labels the location and type of objects within an image—for example, detecting and highlighting people, vehicles, or other targets in real time. This makes it especially suitable for resource-constrained MCU devices. |
| ML_YOLO | Object Detection | Yolov8-nano/YoloX-nano/Yolo Fastest v1.1 | ✔️ | ❌ | ✔️ | The object detection tool using the YoloX-nano model, available through the NuEdgeWise GUI, provides lightweight and efficient real-time object detection capabilities. It is especially suitable for use in embedded systems such as cameras and smart doorbells, accurately and quickly identifying objects and their locations within the scene. |
| ML_Gearbox_Fault_Diagnosis | Anomaly Detection | DNN/Autoencoder | ✔️ | ✔️ | ✔️ | Anomaly Detection, enabled by the graphical tool provided by NuEdgeWise, helps users train models to identify abnormal or unusual conditions—such as unexpected vibrations or noises in mechanical equipment. Even beginners can easily build effective anomaly detection solutions using this intuitive interface. |
| ML_VWW | Visual Wake Words | Small MobileNet RGB/gray | ✔️ | ✔️ | ✔️ | In the microcontroller vision use-case, the objective is to identify whether a person (or any other object of interest) is present in an image. |
Related Document Downloads
- Image Classification
- G-Sensor Gesture Sensing
- Object Detection YoloX-nano
- Keyword Spotting
- Yolo-Fastest(deprecated)