YOLOv8-Segmentation: Optimized for Qualcomm Devices
Ultralytics YOLOv8 is a machine learning model that predicts bounding boxes, segmentation masks and classes of objects in an image.
This is based on the implementation of YOLOv8-Segmentation found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up 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 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-Segmentation on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Input resolution: 640x640
- Model checkpoint: YOLOv8N-Seg
- Model size (float): 13.2 MB
- Model size (w8a16): 3.91 MB
- Number of output classes: 80
- Number of parameters: 3.43M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| YOLOv8-Segmentation | ONNX | float | Snapdragon® X2 Elite | 3.86 ms | 17 - 17 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Snapdragon® X Elite | 7.136 ms | 17 - 17 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.945 ms | 17 - 238 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 13.058 ms | 17 - 223 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 9.403 ms | 15 - 19 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 8.22 ms | 12 - 20 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Qualcomm® QCS8450 | 13.058 ms | 17 - 223 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 9.12 ms | 15 - 18 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 7.136 ms | 17 - 17 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 4.732 ms | 13 - 203 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Elite Mobile | 4.732 ms | 13 - 203 MB | NPU |
| YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.276 ms | 13 - 225 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® X2 Elite | 2.886 ms | 5 - 5 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® X Elite | 4.928 ms | 5 - 5 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 3.428 ms | 5 - 212 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 9.328 ms | 5 - 201 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 6.399 ms | 5 - 16 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 17.139 ms | 1 - 183 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.655 ms | 5 - 6 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8775P | 6.523 ms | 0 - 180 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8650P | 6.523 ms | 0 - 180 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8255P | 6.523 ms | 0 - 180 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8450 | 9.328 ms | 5 - 201 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 6.252 ms | 5 - 15 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 4.928 ms | 5 - 5 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 2.82 ms | 5 - 188 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA7255P | 17.139 ms | 1 - 183 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8295P | 9.435 ms | 0 - 166 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 2.82 ms | 5 - 188 MB | NPU |
| YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.06 ms | 5 - 200 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 3.011 ms | 0 - 108 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 8.633 ms | 4 - 89 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 5.929 ms | 4 - 24 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 16.293 ms | 4 - 80 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.073 ms | 0 - 4 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8775P | 6.011 ms | 4 - 85 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8650P | 6.011 ms | 4 - 85 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8255P | 6.011 ms | 4 - 85 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8450 | 8.633 ms | 4 - 89 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 5.725 ms | 4 - 23 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 2.263 ms | 0 - 94 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA7255P | 16.293 ms | 4 - 80 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8295P | 8.632 ms | 4 - 60 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Elite Mobile | 2.263 ms | 0 - 94 MB | NPU |
| YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.888 ms | 0 - 92 MB | NPU |
License
- The license for the original implementation of YOLOv8-Segmentation can be found here.
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.