ailia-models-tflite

Quantized tflite models for ailia TFLite Runtime

About ailia TFLite Runtime

ailia TF Lite Runtime is a TensorFlow Lite compatible inference engine. Written in C99, it supports inference in Non-OS and RTOS. It also supports high-speed inference using Intel MKL on a PC, and operates 29 times faster than the official TensorFlow Lite.

Install

Get the ailia TF Lite Runtime package from ax Inc. Run the following command.

cd ailia_tflite_runtime/python
python3 bootstrap.py
pip3 install .

Models

Face detection

	Model	Reference	Exported From	Netron
	BlazeFace	PINTO_model_zoo	TensorFlow	Netron

Face recognition

	Model	Reference	Exported From	Netron
	Face Mesh	PINTO_model_zoo	TensorFlow	Netron

Hand recognition

	Model	Reference	Exported From	Netron
	Blaze Hand	PINTO_model_zoo	TensorFlow	Netron

Image classification

	Model	Reference	Exported From	Netron
	MobileNet	MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications	Keras	Netron
	MobileNetV2	MobileNetV2: Inverted Residuals and Linear Bottlenecks	Keras	Netron
	ResNet50	tf.keras.applications.resnet50.ResNet50	Keras	Netron

Image segmentation

	Model	Reference	Exported From	Netron
	DeepLabv3+	PINTO_model_zoo	TensorFlow	Netron

Object detection

	Model	Reference	Exported From	Netron
	MobileNetV2-SSDLite	PINTO_model_zoo	TensorFlow	Netron
	YOLOv3 tiny	tensorflow-yolov4-tflite	TensorFlow	Netron

Options

You can benchmark with the -b option. You can use the official TensorFlow Lite with the --tflite option.