Edge Impulse Experts / AI-based Aquatic Chemical Water Quality Testing
This is your Edge Impulse project. From here you acquire new training data, design impulses and train models.
About this project
This object detection (NVIDIA TAO RetinaNet) model detects water pollution levels based on the applied chemical water quality tests (color-coded):
- dangerous
- polluted
- sterile
After building my RetinaNet object detection model, I deployed my model as a fully optimized and customizable Linux (AARCH64) application (.eim) and uploaded it to UNIHIKER. Also, I developed a user interface (Tkinter-based) from scratch to allow the user to capture image samples effortlessly. UNIHIKER also sends push notifications via the given Telegram bot so as to inform the user of the model detection results by transferring the modified resulting images after running successful inferences.
Download block output
| Title | Type | Size | |
|---|---|---|---|
| Image training data | NPY file | 60 windows | |
| Image training labels | JSON file | 60 windows | |
| Image testing data | NPY file | 9 windows | |
| Image testing labels | JSON file | 9 windows | |
| Object detection model | TensorFlow Lite (float32) | 24 MB | |
| Object detection model | TensorFlow Lite (int8 quantized) | 6 MB | |
| Object detection model | ONNX model | 24 MB | |
| Object detection model | NVIDIA TAO NMS attributes | 379 Bytes | |
| Object detection model | NVIDIA TAO original TensorRT model | 24 MB | |
| Object detection model | Model evaluation metrics (JSON file) | 3 KB | |
| Object detection model | TensorFlow SavedModel | 57 MB |
Clone project
Run this model
Summary
Data collected
69 itemsProject info
| Project ID | 368609 |
| Project version | 1 |
| License | Apache 2.0 |
