Available Targets

Here is the list of all the available targets of the dAIEdge-VLab and their inference engine(s).

Last update : 21.07.2026 at 22:35:11

Target Name	runtime(s)	Description
Jetson Orin Nano	ort tflite trt	Jetson Orin Nano Developer Kit is an official NVIDIA development kit based on the Jetson Orin Nano 8GB module that delivers up to 40 TOPS of AI performance.
JetsonAGXOrin	ort	The NVIDIA Jetson AGX Orin is a high-performance, server-class AI supercomputer designed for next-generation autonomous machines and edge computing, delivering up to 275 TOPS of AI processing power in a compact, energy-efficient form factor.
Kendryte K210	nncase	The Kendryte K210 is a dual-core 64-bit RISC-V SoC designed for low-power edge AI, featuring 8 MB on-chip SRAM (≈6 MB general + 2 MB “AI SRAM” for feature maps), a fixed-function KPU CNN accelerator (1x1/3x3 conv, BN, ReLU, pooling). Models are compiled to .kmodel via NNCase from Keras, TFLite or ONNX.
Kria KV260	hls4ml secda-tflite	The AMD Kria KV260 is based on AMD Zynq UltraScale+ MPSoC architecture paired with 4 GB of DDR4 memory. It features a quad-core ARM Cortex-A53 processor, a Mali-400 MP2 GPU, and an FPGA fabric. HLS4ML framework efficient and fast translation of machine learning models from open-source packages (like Keras and PyTorch) for training machine learning algorithms to high level synthesis (HLS) code that can then be transpiled to run on an FPGA.
NXP LPC55S69	tflite	High Efficiency Arm® Cortex®-M33-Based Microcontroller from NXP.
Qualcomm RB3 Gen2	tflite	The Qualcomm RB3 Gen 2, based on the QCS6490 processor, is an IoT development kit designed for high-performance computing.
Raspberry Pi 4B	ort tflite aidge	Raspberry Pi 4b features theBroadcom BCM2711, Quad core Cortex-A72 (ARM v8) 64-bit SoC @ 1.8GHz. The available Raspberry Pi 4b has 4G of RAM.
Raspberry Pi 5	ort tflite aidge	Raspberry Pi 5 features the Broadcom BCM2712 quad-core Arm Cortex A76 processor @ 2.4GHz, making it up to three times faster than the previous generation. The available Raspberry Pi 5 has 4G of RAM.
Speck Dev Kit	speck	Speck™ is a fully event-driven neuromorphic vision SoC. As it requires a dataset for benchmarking due to its neuromorphic nature, please use the dAIEdge-VLab Python Client instead of the Web user interface. Refer to the user documentation of the VLAb under the [dAIEdge-VLab Targets / Speck] Section for more informations. Benchmark with web UI uses a default NMNIST dataset as generated in the example documentation.
SpeckSim	speck	SpeckSim is a simulator for the Speck hardware device from SynSense that runs spiking neural networks (SNNs). Upload a torch.nn.Sequential model (saved as a .pth file) composed only of Linear or Conv2d layers, trained locally to solve the MNIST task (input_dim = 28*28, output_dim=10). Then, evaluate your model’s performance (accuracy) when executed as a spiking neural network on SpeckSim.
STM32H7A3	stedgeai	High-performance and DSP with DP-FPU, Arm Cortex-M7 MCU with 1MBytes of Flash memory, 1376 KB SRAM, 280 MHz CPU, L1 cache, graphic accelerations, external memory interfaces, SMPS and large set of peripherals.
STM32L4R9	stedgeai tflite x-cube-ai	STM32L4R9 device extend the ultra-low-power portfolio and performance with an Arm® Cortex®-M4 core with DSP and floating-point unit (FPU) at 120 MHz.
STM32MP257	tflite	The STM32MP257 offers advanced edge AI capabilities with its NPU accelerator and provides flexibility in running AI applications on either the CPU, GPU, or NPU. Note that the benchmarks performed by the dAIEdge-VLab are run on the NPU.

Configurations

Here is the list of all the available configurations that can be used with the dAIEdgeVLab Python API. The values in the column Target msut be used for the parameter target in the startBenchmark methode. Same for the parameter runtime.

ID	Target name	`Target`	`Runtime`
1	Jetson Orin Nano	jetsonorinnano	ort
2	Jetson Orin Nano	jetsonorinnano	tflite
3	Jetson Orin Nano	jetsonorinnano	trt
4	JetsonAGXOrin	agx	ort
5	Kendryte K210	k210	nncase
6	Kria KV260	kria-kv260	hls4ml
7	Kria KV260	kria-kv260	secda-tflite
8	NXP LPC55S69	lpc55s69jbd100	tflite
9	Qualcomm RB3 Gen2	qualcommrb3gen2	tflite
10	Raspberry Pi 4B	rpi4b	ort
11	Raspberry Pi 4B	rpi4b	tflite
12	Raspberry Pi 4B	rpi4b	aidge
13	Raspberry Pi 5	rpi5	ort
14	Raspberry Pi 5	rpi5	tflite
15	Raspberry Pi 5	rpi5	aidge
16	Speck Dev Kit	speck_dev_kit	speck
17	SpeckSim	specksim	speck
18	STM32H7A3	stm32h7a3	stedgeai
19	STM32L4R9	stm32l4r9	stedgeai
20	STM32L4R9	stm32l4r9	tflite
21	STM32L4R9	stm32l4r9	x-cube-ai
22	STM32MP257	stm32mp257	tflite

Use example

Here is an example for the configuration 2:

api = dAIEdgeVLabAPI("setup.yaml")

# Start a benchmark for a given target, runtime and model
benchmark_id = api.startBenchmark(
    target = "jetsonorinnano", 
    runtime = "tflite", 
    model_path = MODEL
    )