tflite-cv-example
TensorFlow Lite, Coral Edge TPU samples (Python/C++, Raspberry Pi/Windows/Linux).
MIT License
Stars
118
Ecosystems:
Python
TensorFlow Lite samples.
About
TensorFlow Lite samples (Python/C++, Raspberry Pi/VisionFive 2/Windows/Linux).
- CPU(XNNPACK) inference
- Coral Edge TPU Delegate
- GPU Delegate
List of samples.
| Name | Language | Description | API | OS |
|---|---|---|---|---|
| Camouflage | Python | Object detection and camouflage objects by PiCamera. | PyCoral | LinuxWindows |
| Classify | Python | Image classifilcation by PiCamera or Video Capture. | TF-LitePyCoral | LinuxWindows |
| CenterNet | PythonC++ | CenterNet on-device with TensorFlow Lite. | TF-Lite | LiuxWindows |
| DeepLab | PythonC++ | Semantic Segmentation using DeepLab v3. | TF-LiteEdgeTPU API | LinuxWindows |
| Object detection | PythonC++VC++ | Object detection by PiCamera or Video Capture. | TF-LitePyCoral | LinuxWindows |
| U-Net MobileNet v2 | Python | Image segmentation model U-Net MobileNet v2. | TF-Lite | LinuxWindows |
| Super resolution | Python | Super resolution using ESRGAN. | TF-Lite | LinuxWindows |
| YOLOX | Python | YOLOX with TensorFlow Lite. | TF-Lite | LinuxWindows |
| DeepLab V3+ EdgeTPUV2 and AutoSeg EdgeTPU | Python | DeepLab V3+ EdgeTPUV2 and AutoSeg EdgeTPU with TensorFlow Lite. | TF-LiteEdgeTPU | LinuxWindows |
| FFNet | C++ | VisionFive 2 TensorFlow Lite GPU Delegate FFNet | TF-LiteGPU delegate | Linux |
Images
| Object detection | Camouflage | DeepLab |
|---|---|---|
 |
 |
 |
| Segmentation | CenterNet | YOLOX |
|---|---|---|
 |
 |
 |
| DeepLab V3+ EdgeTPUV2 and AutoSeg EdgeTPU | VisionFive 2 TensorFlow Lite GPU DelegateFFNet46NS CCC Mobile Pre-Down Fused-Argmax | VisionFive 2 TensorFlow Lite GPU DelegateEfficientDet-Lite0 |
|---|---|---|
YouTube Link
|
YouTube Link
|
YouTube Link
|
Environment
- Coral Edge TPU USB Accelerator
- Raspberry Pi (3 B+ / 4) + PiCamera or UVC Camera
- Dev Board
- VisionFive 2
- x64 PC(Windows or Linux) + Video file or UVC Camera
- Python3
Installation
- OpenCV with OpenCV's extra modules(3.4.5 or higher)
- TensorFlow Lite Runtime (Python quickstart).
- Edge TPU Python library (Get started with the USB Accelerator).
Reference
Badges
Extracted from project README