---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: object-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolov8_det/web-assets/model_demo.png)

# YOLOv8-Detection: Optimized for Qualcomm Devices

Ultralytics YOLOv8 is a machine learning model that predicts bounding boxes and classes of objects in an image.

This is based on the implementation of YOLOv8-Detection found [here](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/models/yolo/detect).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov8_det) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov8_det) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

See our repository for [YOLOv8-Detection on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov8_det) for usage instructions.


## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- Model checkpoint: YOLOv8-N
- Input resolution: 640x640
- Number of parameters: 3.18M
- Model size (float): 12.2 MB
- Model size (w8a8): 3.25 MB
- Model size (w8a16): 3.60 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| YOLOv8-Detection | ONNX | float | Snapdragon® X Elite | 5.576 ms | 5 - 5 MB | NPU
| YOLOv8-Detection | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.466 ms | 5 - 214 MB | NPU
| YOLOv8-Detection | ONNX | float | Qualcomm® QCS8550 (Proxy) | 5.398 ms | 5 - 12 MB | NPU
| YOLOv8-Detection | ONNX | float | Qualcomm® QCS9075 | 7.069 ms | 5 - 7 MB | NPU
| YOLOv8-Detection | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.881 ms | 0 - 154 MB | NPU
| YOLOv8-Detection | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.381 ms | 0 - 153 MB | NPU
| YOLOv8-Detection | ONNX | w8a16 | Snapdragon® X Elite | 4.93 ms | 2 - 2 MB | NPU
| YOLOv8-Detection | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.918 ms | 0 - 154 MB | NPU
| YOLOv8-Detection | ONNX | w8a16 | Qualcomm® QCS6490 | 344.267 ms | 65 - 70 MB | CPU
| YOLOv8-Detection | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 4.789 ms | 0 - 142 MB | NPU
| YOLOv8-Detection | ONNX | w8a16 | Qualcomm® QCS9075 | 5.779 ms | 2 - 5 MB | NPU
| YOLOv8-Detection | ONNX | w8a16 | Qualcomm® QCM6690 | 169.623 ms | 67 - 76 MB | CPU
| YOLOv8-Detection | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 2.278 ms | 1 - 131 MB | NPU
| YOLOv8-Detection | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 156.125 ms | 71 - 80 MB | CPU
| YOLOv8-Detection | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2.04 ms | 0 - 133 MB | NPU
| YOLOv8-Detection | ONNX | w8a8 | Snapdragon® X Elite | 1.771 ms | 2 - 2 MB | NPU
| YOLOv8-Detection | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.225 ms | 0 - 136 MB | NPU
| YOLOv8-Detection | ONNX | w8a8 | Qualcomm® QCS6490 | 60.772 ms | 18 - 27 MB | CPU
| YOLOv8-Detection | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.809 ms | 0 - 5 MB | NPU
| YOLOv8-Detection | ONNX | w8a8 | Qualcomm® QCS9075 | 2.203 ms | 0 - 3 MB | NPU
| YOLOv8-Detection | ONNX | w8a8 | Qualcomm® QCM6690 | 41.45 ms | 19 - 30 MB | CPU
| YOLOv8-Detection | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.925 ms | 0 - 116 MB | NPU
| YOLOv8-Detection | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 31.536 ms | 21 - 32 MB | CPU
| YOLOv8-Detection | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.805 ms | 0 - 117 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Snapdragon® X Elite | 4.026 ms | 5 - 5 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 2.698 ms | 0 - 246 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 13.578 ms | 0 - 164 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 3.644 ms | 5 - 114 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® SA8775P | 5.245 ms | 1 - 166 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® QCS9075 | 4.895 ms | 5 - 11 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 8.113 ms | 5 - 190 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® SA7255P | 13.578 ms | 0 - 164 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Qualcomm® SA8295P | 7.924 ms | 0 - 159 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.986 ms | 5 - 205 MB | NPU
| YOLOv8-Detection | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.729 ms | 5 - 202 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Snapdragon® X Elite | 3.701 ms | 2 - 2 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 2.246 ms | 2 - 73 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 8.736 ms | 2 - 6 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 6.661 ms | 1 - 52 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 3.34 ms | 2 - 14 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® SA8775P | 3.903 ms | 0 - 54 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 3.711 ms | 1 - 5 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 20.961 ms | 2 - 169 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 4.012 ms | 2 - 74 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® SA7255P | 6.661 ms | 1 - 52 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Qualcomm® SA8295P | 4.569 ms | 0 - 48 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 1.538 ms | 2 - 55 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 4.664 ms | 2 - 171 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.254 ms | 2 - 60 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.607 ms | 1 - 1 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.965 ms | 1 - 53 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 3.619 ms | 3 - 6 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 3.222 ms | 1 - 38 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.388 ms | 1 - 3 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.784 ms | 1 - 39 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 1.684 ms | 1 - 4 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 8.472 ms | 1 - 155 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 1.658 ms | 1 - 53 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® SA7255P | 3.222 ms | 1 - 38 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Qualcomm® SA8295P | 2.216 ms | 0 - 35 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.695 ms | 1 - 36 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.526 ms | 1 - 37 MB | NPU
| YOLOv8-Detection | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.584 ms | 0 - 39 MB | NPU
| YOLOv8-Detection | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.407 ms | 0 - 174 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 13.069 ms | 0 - 102 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 3.267 ms | 0 - 4 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® SA8775P | 4.919 ms | 0 - 105 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® QCS9075 | 4.599 ms | 0 - 14 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 6.245 ms | 0 - 74 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® SA7255P | 13.069 ms | 0 - 102 MB | NPU
| YOLOv8-Detection | TFLITE | float | Qualcomm® SA8295P | 7.17 ms | 0 - 49 MB | NPU
| YOLOv8-Detection | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.822 ms | 0 - 109 MB | NPU
| YOLOv8-Detection | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.411 ms | 0 - 110 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.908 ms | 0 - 51 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® QCS6490 | 3.301 ms | 0 - 7 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 57.151 ms | 8 - 40 MB | GPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.332 ms | 0 - 2 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® SA8775P | 1.697 ms | 0 - 40 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® QCS9075 | 1.673 ms | 0 - 7 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® QCM6690 | 8.417 ms | 0 - 153 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 1.619 ms | 0 - 51 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® SA7255P | 57.151 ms | 8 - 40 MB | GPU
| YOLOv8-Detection | TFLITE | w8a8 | Qualcomm® SA8295P | 2.129 ms | 0 - 35 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.659 ms | 0 - 33 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.458 ms | 0 - 34 MB | NPU
| YOLOv8-Detection | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.587 ms | 0 - 40 MB | NPU

## License
* The license for the original implementation of YOLOv8-Detection can be found
  [here](https://github.com/ultralytics/ultralytics/blob/main/LICENSE).

## References
* [Ultralytics YOLOv8 Docs: Object Detection](https://docs.ultralytics.com/tasks/detect/)
* [Source Model Implementation](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/models/yolo/detect)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).