Arm/yolo12l-int8-xnnpack-executorch-raspberrypi5

YOLO12-L optimized for Arm-based Edge Linux systems

An INT8-quantized version of YOLO12-L for object detection, exported to ExecuTorch (.pte) and optimized for Arm-based Edge Linux systems.

Summary

This repository contains an Arm-optimized version of YOLO12-L for object detection, quantized to INT8 via static post-training quantization — per-channel symmetric weights, per-tensor affine activations. The model is provided in ExecuTorch (.pte) format, targeting Edge Linux systems.

YOLO12-L is an attention-centric single-stage detector for the 80 COCO categories. Its area-attention blocks introduce attention math into the graph, so the quantization recipe keeps a small set of layers in FP32 where a shared per-tensor scale would otherwise collapse a tensor's dynamic range.

This version is intended to demonstrate efficient inference on Arm-based platforms while preserving the original model's intended behavior. Arm has evaluated this model on COCO 2017 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, Linux Raspberry Pi OS 64-bit, based on Debian 13 "Trixie")
Primary performance result	p50 latency 2650.581 ms (1.65x faster than baseline)
Accuracy result	mAP@0.5:0.95 52.29%
Size / memory result	33.809 MB (3.00x smaller), peak memory 268.141 MB

Original model

Field	Value
Original model	YOLO12-L
Original source	GitHub
Original developer	Ultralytics
Original model card	YOLO12
Original license	AGPL-3.0

Model files

File	Description
yolo12l_executorch_optimized.pte	Arm-optimized INT8 model for deployment
example.py	Minimal inference example
pyproject.toml	Pinned runtime dependencies for example.py, resolved with uv
uv.lock	Locked dependency resolution for pyproject.toml
config.yaml	Model I/O contract used by the example
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	Cortex-A76, 4 cores @ 2.4 GHz
System memory	8 GB
OS	Linux, Raspberry Pi OS 64-bit, based on Debian 13 "Trixie"
Runtime	ExecuTorch 1.1.0
Execution backend	CPU (XNNPACK, KleidiAI)
Precision	INT8, static PTQ — per-channel symmetric weights, per-tensor affine activations
Batch size	1
Input resolution	640x640
Threads	1
Runs / warmup	100 / 10

Performance results

Metric	Original / baseline	Arm-optimized	Improvement
Model size (MB)	101.32	33.809	3.00x smaller
End-to-end latency p50 (ms)	4374.922	2650.581	1.65x faster
End-to-end latency p90 (ms)	4381.052	2652.596	1.65x faster
End-to-end latency p99 (ms)	4385.159	2654.184	1.65x faster
Model load time (ms)	126.273	69.875	1.81x faster
Time to first inference (ms)	4426.724	2705.997	1.64x faster
Peak memory (MB)	335.281	268.141	1.25x less
Average memory (MB)	327.391	216.141	1.51x less
Frames per second	0.229	0.377	1.65x

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	COCO 2017
Split	val2017
Sample count	5000
Metric(s)	mAP@0.5:0.95, mAP@0.5, mAP@0.75
Runtime	ExecuTorch 1.1.0 (CPU, XNNPACK, KleidiAI)

Accuracy results

Metric	Original / baseline	Arm-optimized	Change
mAP@0.5:0.95 (%)	52.77	52.29	-0.48 pp
mAP@0.5 (%)	69.35	68.98	-0.37 pp
mAP@0.75 (%)	57.41	56.89	-0.52 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
Quantization	Yes	INT8 PTQ-static — per-channel symmetric weights, per-tensor affine activations; calibrated on 300 COCO 2017 samples; selected layers were skipped from quantization to keep accuracy within acceptable ranges
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

Dependencies are declared in pyproject.toml, which ships with this repository. Resolve and install them into a local virtual environment with uv:

uv python install
uv sync --frozen

Run the example

uv run example.py

Expected input

Property	Value
Shape	[1, 3, 640, 640]
Dtype	float32
Range	[0.0, 1.0], RGB, no mean/std normalization
Preprocessing	scale pixels to [0.0, 1.0], then aspect-preserving letterbox to 640x640 — uniform scale min(640 / width, 640 / height), bilinear resize with align_corners=False, center-pasted onto a canvas filled with gray 114 / 255

Expected output

Property	Value
Shape	[1, 84, N]
Format	84 = 4 box values (x_center, y_center, width, height, in the 640x640 model frame) + 80 class scores, no objectness channel
Postprocessing	transpose to [1, N, 84], sigmoid only if class scores fall outside [0, 1], score = max and label = argmax over the 80 class scores, confidence filter, center-to-corner conversion, per-class NMS at IoU 0.7, then the inverse letterbox back to original-image coordinates

Intended use

This model is intended for developers evaluating object detection 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 COCO 2017 val2017 and may not generalize to all domains.
• 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

The input size is fixed at 640x640. Images of other sizes must be letterboxed as described above, and predictions un-mapped through the same transform — an anisotropic resize distorts the aspect ratio and shifts every predicted box.

The quantization recipe keeps the detection head, the stem convolution, and the consumers of the surviving concatenation nodes in FP32. Concatenations that fuse tensors of very different dynamic range — box coordinates spanning 0-640 alongside class scores in [0, 1] — collapse the smaller range under a shared per-tensor scale, so those nodes and the convolutions reading them are held at higher precision while every other convolution stays INT8. Calibration used 300 images from COCO 2017, disjoint from the val2017 split used for evaluation.

• Sample input: sample_input.jpg is derived from Living room (Unsplash) by Jarosław Ceborski, via Wikimedia Commons (CC0 1.0).

About this version

Original Model: YOLO12-L by Ultralytics - 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 AGPL-3.0.

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.

Use of Ultralytics models

Ultralytics provides Ultralytics YOLO software and models available through the Arm AI Portal and/or Arm's Hugging Face Organization under the GNU Affero General Public License v3.0 ("AGPL-3.0"), unless you have entered into a separate written license agreement with Ultralytics. Accessing, downloading, or retraining these materials through Arm AI Portal does not grant you an Ultralytics Enterprise License or any other proprietary Ultralytics license.

AGPL-3.0 requires you to release the complete source code of any application that uses Ultralytics YOLO, including applications made available over a network, under the same license. If you are embedding YOLO in a commercial product, internal tool, or production deployment and cannot open-source your code, you need an Ultralytics Enterprise License.

You are responsible for determining which applies to your use. Terms and enterprise licensing options are available at ultralytics.com.