Arm/shufflenet-v2-x1-0-int8-xnnpack-executorch

ShuffleNetV2 x1.0 optimized for Arm-based Edge Linux

ShuffleNetV2 x1.0 image classification quantized to INT8 with static post-training quantization (PTQ) and exported as an ExecuTorch .pte model for Arm-based Edge Linux systems.

Summary

This repository contains an Arm-optimized version of shufflenet_v2_x1_0 for image classification. The model is provided in ExecuTorch .pte (ExecuTorch runtime), targeting Edge Linux systems.

This version demonstrates efficient inference on Arm-based platforms while preserving the original model's intended behavior. Arm evaluated this model on ImageNet-1k and measured performance on a representative evaluation target.

Key results

Area	Result
Model format	ExecuTorch .pte
Target device class	Edge Linux
Reference device	Raspberry Pi 5 (Cortex-A76, Raspberry Pi OS 64-bit, based on Debian 13 "Trixie")
Primary performance result	3.63 ms p50 latency (275.39 FPS)
Accuracy result	Top-1 68.76% / Top-5 88.01%
Size / memory result	2.40 MB (.pte), 3.64x smaller than FP32

Original model

Field	Value
Original model	shufflenet_v2_x1_0
Original source	torchvision/models/shufflenetv2.py
Original developer	Megvii Inc. for the architecture; PyTorch TorchVision for the pretrained model
Original model card	torchvision.models.shufflenet_v2_x1_0
Original license	BSD-3-Clause

Model files

File	Description
shufflenet_v2_x1_0_raspberry_executorch_optimized.pte	Arm-optimized INT8 model for deployment
example.py	Minimal inference example
pyproject.toml	Pinned runtime dependencies and deployment metadata
uv.lock	Locked dependency versions, sources, and hashes for uv sync --frozen
config.yaml	Model I/O contract used by the example
sample_input.jpg	Example input
benchmarks/	FP32 baseline and Arm-optimized benchmark records

Performance

Performance was measured on the reference configuration below. Results are intended to make the optimization reproducible but do not guarantee identical performance on every Arm-based system.

Reference configuration

Field	Value
Device / platform	Raspberry Pi 5
CPU / accelerator	Cortex-A76 (4 cores @ 2.4 GHz), CPU
OS	Raspberry Pi OS 64-bit, based on Debian 13 "Trixie"
Runtime	ExecuTorch 1.1.0
Backend / delegate	XNNPACK + KleidiAI
Batch size	1
Precision	INT8 PTQ-static (symmetric per-channel weights, asymmetric per-tensor affine activations)
Runs	10 warmup + 100 measured

Performance results

Metric	Original / baseline	Arm-optimized	Improvement
p50 latency	6.89 ms	3.63 ms	1.90x
p90 latency	6.95 ms	3.78 ms	1.84x
p99 latency	8.42 ms	5.27 ms	1.60x
Frames per second	145.22 FPS	275.39 FPS	1.90x
Model size	8.76 MB	2.40 MB	3.64x smaller
Peak memory	16.53 MB	9.22 MB	1.79x less

Accuracy

Accuracy was evaluated using the same preprocessing, input resolution, and evaluation protocol described below. Where possible, the optimized model is compared against the original model under the same evaluation conditions.

Evaluation setup

Field	Value
Dataset	ImageNet-1k
Split	val
Number of samples	50000
Metric(s)	Top-1 Accuracy, Top-5 Accuracy
Evaluation runtime	ExecuTorch

Accuracy results

Metric	Original / baseline	Arm-optimized	Change
Top-1 Accuracy	69.36%	68.76%	-0.60 pp
Top-5 Accuracy	88.31%	88.01%	-0.30 pp

Accuracy was measured using the evaluation setup described above. Users should re-evaluate the model on their own data before production use.

Arm optimization approach

Arm optimized this model for efficient inference on Arm-based platforms using a hardware-aware conversion and validation flow.

For this release, Arm used:

Optimization area	Applied?	Notes
Model conversion	Yes	Converted to ExecuTorch .pte through the PT2E export path
Quantization	Yes	Static INT8 PTQ with symmetric per-channel weights and asymmetric per-tensor affine activations, calibrated on 1,000 randomly selected ImageNet-1k samples. No layers are excluded from quantization; the first convolution is quantized like the rest
Runtime/backend selection	Yes	XNNPACK + KleidiAI delegate
Graph/runtime compatibility updates	Yes	Performed as part of the ExecuTorch export pipeline
Accuracy validation	Yes	Compared against the original model or published baseline
Performance validation	Yes	Measured on the reference Arm platform

The goal of this process is to improve deployment characteristics such as latency, memory use, model size, and runtime compatibility while preserving the model's intended behavior. Detailed conversion scripts, calibration configuration, or backend-specific implementation details may be provided separately where appropriate.

Using this model

Install dependencies

From this model directory:

uv python install
uv sync --frozen

Run the example

uv run example.py

The example writes its top-five predictions to predictions.json and an annotated copy of the input to sample_output.jpg, both beside the script; the predictions.json and sample_output.jpg in this repository record the expected result for the committed sample_input.jpg. Run it from a copy of the directory if you want to keep those reference files untouched.

Deployment requirements

The validated deployment target is a Raspberry Pi 5 (Cortex-A76, aarch64) running Raspberry Pi OS 64-bit, based on Debian 13 "Trixie" (glibc 2.41). Inference runs on the CPU through ExecuTorch 1.1.0 and requires the XnnpackBackend runtime capability, which was confirmed on the device through the ExecuTorch backend registry. No additional Apt packages are required on Raspberry Pi OS or Ubuntu.

Expected input

Property	Value
Input shape	[1, 3, 224, 224]
Input type	float32
Input range	[0.0, 1.0]
Preprocessing	Resize 256, center-crop 224×224, ImageNet mean/std normalization

Expected output

Property	Value
Output shape	[1, 1000]
Output type	float32 (raw class logits)
Postprocessing	softmax, top-5

Intended use

This model is intended for developers evaluating image-classification workloads on Arm-based platforms. It is suitable as a reference implementation for benchmarking, prototyping, and integration exploration.

Limitations

• Performance depends on the target device, runtime version, backend/delegate support, memory configuration, and system load.
• Accuracy was evaluated on ImageNet-1k val and may not generalize to all domains.
• The exported model has a fixed 224×224 input resolution.
• This release preserves the original model's intended task and behavior, but users should validate it for their own application, data, and deployment environment.
• This repository is not a replacement for the original model documentation.

Additional notes

• ShuffleNetV2 relies on channel split, depthwise convolution, and channel shuffle operations; the exported graph keeps those structural operations outside the delegated INT8 subgraphs.
• The example reads its 1000 ImageNet class names from TorchVision's bundled category list, so the names it prints are the short primary forms.
• Sample input: sample_input.jpg is derived from Samoyed on Beach by Appleinfl, via Wikimedia Commons (public domain).

About this version

Original Model: shufflenet_v2_x1_0 by Megvii Inc. for the architecture; PyTorch TorchVision for the pretrained model - Repository

Optimization/conversion: Arm-Optimized version for execution on Arm-based platforms.

Converted/optimized by: Arm

License: The Original Model and the Optimized Model are subject to BSD-3-Clause.

This repository contains a converted or optimized version of the Original Model (the “Optimized Model”). The Original Model has been converted or optimized as described above for execution on Arm-based platforms.

No retraining or fine-tuning of the Original Model was performed as part of the conversion or optimization. The conversion or optimization was not intended to change the Original Model’s behavior or intended use.

Original Model and Documentation

For information about the Original Model, including its development, training data, intended uses, limitations and other relevant information, please refer to the Original Model repository. Information in that repository was provided by the original developer or other third parties and, unless expressly stated otherwise, has not been independently verified by Arm.

Licenses and Third-Party Terms

Use of the Original Model and the Optimized Model is subject to the applicable licenses, usage restrictions and other terms identified above and in the relevant repositories. Publication of the Optimized Model does not grant any rights beyond those provided under the applicable license terms.

You are responsible for reviewing those terms and ensuring that your use of the Original Model and the Optimized Model is permitted.

Purpose of this Release

The Optimized Model is provided as a reference implementation to demonstrate and evaluate execution and performance on Arm-based systems. It is not a production-ready or supported solution.

Arm’s publication of the Optimized Model does not constitute an endorsement or certification of the Original Model or a representation that the Optimized Model is suitable for production use or any particular purpose.

To the fullest extent permitted by applicable law (i) the Optimized Model is provided “as is.” Arm makes no representations or warranties that the Original Model, the Optimized Model or their outputs are accurate, safe, secure, non-infringing, legally compliant, suitable for production use or fit for any particular purpose; and (ii) Arm will not be liable for any loss or damage arising from or in connection with the Optimized Model, its use or its outputs.

You are responsible for independently evaluating the Optimized Model, its outputs and its suitability for your intended use, including compliance with applicable legal, regulatory, safety and security requirements.

Arm does not commit to provide ongoing support, maintenance or updates for the Optimized Model. Any use of or reliance on the Optimized Model or its outputs is at your own risk.