Convert

You can use LaunchX converter to automatically convert the AI model’s framework to the target framework.

Conversion case

1. ONNX to TensorRT

2. ONNX to TensorFlow Lite

3. ONNX to OpenVINO

4. TensorFlow-Keras to TensorFlow Lite

Compatible model

The input layer of the uploaded model should be as follows.

• Only single-input models are supported.

• The four-dimensional array structure of images should be organized Batch, Number of Channels, Height, and Width.

1. Batch size: The number of combined input datasets that the model processes simultaneously.

2. Channel: 3 for RGB or BGR and 1 for Grayscale.

3. Input size: In computer vision tasks, input size refers to the size of the input images.

ONNX to TensorRT

• Supported JetPack-ONNX version

Target Device	JetPack version	Input data type	Batch size	Channel	Input size	Output data type
NVIDIA Jetson Nano	4.6, 4.4.1	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16
NVIDIA Jetson Xavier NX	5.0.2, 4.6	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16
NVIDIA Jetson TX2	4.6	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16
NVIDIA Jetson AGX Xavier	4.6	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16
NVIDIA Jetson AGX Orin	5.0.1	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16
NVIDIA Jetson Orin Nano	6.0	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16
NVIDIA T4	None	FP32	1~4 (Static), Dynamic	1~4	height, width	FP16

ONNX to TFlite

Input data type	Batch size	Channel	Input size	Output data type
FP32	1~4 (Static), Dynamic	1~4	height, width	FP16, INT8

ONNX to OpenVino

Input data type	Batch size	Channel	Input size	Output data type
FP32	1~4 (Static), Dynamic	1~4	height, width	FP16

TensorFlow to TensorFlowLite

Input data type	Batch size	Channel	Input size	Output data type
FP32	1~4 (Static), Dynamic	1~4	height, width	FP16, INT8