yolov8n_seg

Yolov8n_seg

Use case : Instance segmentation

Model description

Yolov8n_seg is a lightweight and efficient 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_seg is implemented in Pytorch by Ultralytics and is quantized in int8 format using tensorflow lite converter.

Network information

Recommended platform

---

Performances

Metrics

Measures are done with default STEdgeAI Core version configuration with enabled input / output allocated option.

[!CAUTION] All YOLOv8 hyperlinks in the tables below link to an external GitHub folder, which is subject to its own license terms: https://github.com/stm32-hotspot/ultralytics/blob/main/LICENSE Please also check the folder's README.md file for detailed information about its use and content: https://github.com/stm32-hotspot/ultralytics/blob/main/examples/YOLOv8-STEdgeAI/README.md

Reference NPU memory footprint based on COCO dataset

|Model | Dataset | Format | Resolution | Series | Internal RAM (KiB)| External RAM (KiB)| Weights Flash (KiB) | STEdgeAI Core version | |----------|------------------|--------|-------------|------------------|------------------|---------------------|-------------------------| | Yolov8n seg per channel | COCO | Int8 | 256x256x3 | STM32N6 | 855 | 0.0 | 3393.42 | 3.0.0 | | Yolov8n seg per channel | COCO | Int8 | 320x320x3 | STM32N6 | 1413.89 | 0.0 | 3435.34 | 3.0.0 |

Reference NPU inference time based on COCO Person dataset

| Model | Dataset | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STEdgeAI Core version | |--------|------------------|--------|-------------|------------------|------------------|---------------------|-------|----------------------|-------------------------| | YOLOv8n seg per channel | COCO-Person | Int8 | 256x256x3 | STM32N6570-DK | NPU/MCU | 31.57 | 29.72 | 3.0.0 | | YOLOv8n seg per channel | COCO-Person | Int8 | 320x320x3 | STM32N6570-DK | NPU/MCU | 41.87 | 22.83 | 3.0.0 |

Reference MPU inference time based on COCO 2017 Person dataset (instance segmentation)

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

** Note: On STM32MP2 devices, per-channel quantized models are internally converted to per-tensor quantization by the compiler using an entropy-based method. This may introduce a slight loss in accuracy compared to the original per-channel models.

Retraining and Integration in a Simple Example

Please refer to the stm32ai-modelzoo-services GitHub here. For instance segmentation, 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 model.

References

<a id="1">[1]</a> T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, and C. L. Zitnick, "Microsoft COCO: Common Objects in Context." European Conference on Computer Vision (ECCV), 2014. Link

<a id="2">[2]</a> Ultralytics, "YOLOv8: Next-Generation Object Detection and Segmentation Model." Ultralytics, 2023. Link