Yolov8n object detection quantized

Use case : `Object detection`

Model description

Yolov8n is a lightweight and efficient object detection model designed for instance segmentation tasks. It is part of the YOLO (You Only Look Once) family of models, known for their real-time object detection capabilities. The "n" in Yolov8n_seg indicates that it is a nano version, optimized for speed and resource efficiency, making it suitable for deployment on devices with limited computational power, such as mobile devices and embedded systems.

Yolov8n is implemented in Pytorch by Ultralytics and is quantized in int8 format using tensorflow lite converter.

Network information

Network information	Value
Framework	TensorFlow Lite
Quantization	int8
Provenance	https://docs.ultralytics.com/tasks/detect/

Networks inputs / outputs

With an image resolution of NxM and K classes to detect:

Input Shape	Description
(1, N, M, 3)	Single NxM RGB image with UINT8 values between 0 and 255

Output Shape	Description
(1, 4+K, F)	FLOAT values Where F = (N/8)^2 + (N/16)^2 + (N/32)^2 is the 3 concatenated feature maps

Recommended Platforms

Platform	Supported	Recommended
STM32L0	[]	[]
STM32L4	[]	[]
STM32U5	[]	[]
STM32H7	[]	[]
STM32MP1	[]	[]
STM32MP2	[x]	[x]
STM32N6	[x]	[x]

Performances

Metrics

Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.

Reference NPU memory footprint based on COCO Person dataset (see Accuracy for details on dataset)

Model	Dataset	Format	Resolution	Series	Internal RAM	Weights Flash	STM32Cube.AI version	STEdgeAI Core version
YOLOv8n per channel	COCO-Person	Int8	192x192x3	STM32N6	697.5	2965.61	10.0.0	2.0.0
YOLOv8n per channel	COCO-Person	Int8	256x256x3	STM32N6	1626	2970.13	10.0.0	2.0.0
YOLOv8n per channel	COCO-Person	Int8	320x320x3	STM32N6	2162.5	2975.99	10.0.0	2.0.0
YOLOv8n per channel	COCO-Person	Int8	416x416x3	STM32N6	2704	2987.52	10.0.0	2.0.0

Reference NPU inference time based on COCO Person dataset (see Accuracy for details on dataset)

Model	Dataset	Format	Resolution	Board	Execution Engine	Inference time (ms)	Inf / sec	STM32Cube.AI version	STEdgeAI Core version
YOLOv8n per channel	COCO-Person	Int8	192x192x3	STM32N6570-DK	NPU/MCU	18.91	52.89	10.0.0	2.0.0
YOLOv8n per channel	COCO-Person	Int8	256x256x3	STM32N6570-DK	NPU/MCU	28.6	34.97	10.0.0	2.0.0
YOLOv8n per channel	COCO-Person	Int8	320x320x3	STM32N6570-DK	NPU/MCU	38.32	26.09	10.0.0	2.0.0
YOLOv8n per channel	COCO-Person	Int8	416x416x3	STM32N6570-DK	NPU/MCU	63.03	15.86	10.0.0	2.0.0

Reference MPU inference time based on COCO Person dataset (see Accuracy for details on dataset)

Model	Format	Resolution	Quantization	Board	Execution Engine	Frequency	Inference time (ms)	%NPU	%GPU	%CPU	X-LINUX-AI version	Framework
YOLOv8n per channel	Int8	256x256x3	per-channel**	STM32MP257F-DK2	NPU/GPU	800 MHz	102.8 ms	11.70	88.30	0	v5.0.0	OpenVX
YOLOv8n per tensor	Int8	256x256x3	per-tensor	STM32MP257F-DK2	NPU/GPU	800 MHz	17.57 ms	86.79	13.21	0	v5.0.0	OpenVX

** To get the most out of MP25 NPU hardware acceleration, please use per-tensor quantization

AP on COCO Person dataset

Dataset details: link , License CC BY 4.0 , Quotation[1] , Number of classes: 80, Number of images: 118,287

Model	Format	Resolution	AP*
YOLOv8n per channel	Int8	192x192x3	56.90 %
YOLOv8n per channel	Int8	256x256x3	62.60 %
YOLOv8n per channel	Int8	320x320x3	66.20 %
YOLOv8n per channel	Int8	416x416x3	68.90 %

* EVAL_IOU = 0.4, NMS_THRESH = 0.5, SCORE_THRESH =0.001

Integration in a simple example and other services support:

Please refer to the stm32ai-modelzoo-services GitHub here. The models are stored in the Ultralytics repository. You can find them at the following link: Ultralytics YOLOv8-STEdgeAI Models.

Please refer to the Ultralytics documentation to retrain the models.

References

[1] “Microsoft COCO: Common Objects in Context”. [Online]. Available: https://cocodataset.org/#download. @article{DBLP:journals/corr/LinMBHPRDZ14, author = {Tsung{-}Yi Lin and Michael Maire and Serge J. Belongie and Lubomir D. Bourdev and Ross B. Girshick and James Hays and Pietro Perona and Deva Ramanan and Piotr Doll{'{a} }r and C. Lawrence Zitnick}, title = {Microsoft {COCO:} Common Objects in Context}, journal = {CoRR}, volume = {abs/1405.0312}, year = {2014}, url = {http://arxiv.org/abs/1405.0312}, archivePrefix = {arXiv}, eprint = {1405.0312}, timestamp = {Mon, 13 Aug 2018 16:48:13 +0200}, biburl = {https://dblp.org/rec/bib/journals/corr/LinMBHPRDZ14}, bibsource = {dblp computer science bibliography, https://dblp.org} }