This project aims to provide a Docker container that facilitates the conversion of YOLOv8 models to TensorFlow Lite (TFLite) format. This conversion is crucial for deploying YOLOv8 models on edge devices, such as google-coral ,smartphones, IoT devices, and embedded systems.
Docker installed on your system.
A trained YOLOv8 model in PyTorch format (.pt).
- Clone this repository to your local machine.
- Place your tflite model in the
model_yolo
folder at the root of the project. - use docker-compose up
.
├── app.py
├── docker-compose.yml
├── Dockerfile
├── model_yolo
│ ├── yolov8s_672.pt
│ └── yolov8s_672_saved_model
│ ├── assets
│ ├── fingerprint.pb
│ ├── metadata.yaml
│ ├── saved_model.pb
│ ├── variables
│ │ ├── variables.data-00000-of-00001
│ │ └── variables.index
│ ├── yolov8s_672_float16.tflite
│ ├── yolov8s_672_float32.tflite
│ ├── yolov8s_672_full_integer_quant_edgetpu.log
│ ├── yolov8s_672_full_integer_quant_edgetpu.tflite
│ ├── yolov8s_672_full_integer_quant.tflite
│ ├── yolov8s_672_int8.tflite
│ └── yolov8s_672_integer_quant.tflite
└── README.md