YOLOv8-Segmentation: Optimized for Mobile Deployment
Real-time object segmentation optimized for mobile and edge by Ultralytics
Ultralytics YOLOv8 is a machine learning model that predicts bounding boxes, segmentation masks and classes of objects in an image.
This model is an implementation of YOLOv8-Segmentation found here.
This repository provides scripts to run YOLOv8-Segmentation on Qualcomm® devices. More details on model performance across various devices, can be found here.
WARNING: The model assets are not readily available for download due to licensing restrictions.
Model Details
- Model Type: Model_use_case.semantic_segmentation
- Model Stats:
- Model checkpoint: YOLOv8N-Seg
- Input resolution: 640x640
- Number of output classes: 80
- Number of parameters: 3.43M
- Model size (float): 13.2 MB
- Model size (w8a16): 3.91 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| YOLOv8-Segmentation | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 17.723 ms | 4 - 75 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 16.601 ms | 5 - 118 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 8.708 ms | 4 - 52 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 10.625 ms | 5 - 49 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 4.871 ms | 0 - 39 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 4.357 ms | 5 - 25 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 6.41 ms | 0 - 106 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 6.686 ms | 4 - 75 MB | NPU | -- |
| YOLOv8-Segmentation | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 6.11 ms | 1 - 112 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 17.723 ms | 4 - 75 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 16.601 ms | 5 - 118 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 4.951 ms | 0 - 38 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 4.333 ms | 6 - 22 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 9.889 ms | 4 - 41 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 8.462 ms | 2 - 36 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 4.886 ms | 0 - 37 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 4.355 ms | 5 - 18 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 6.686 ms | 4 - 75 MB | NPU | -- |
| YOLOv8-Segmentation | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 6.11 ms | 1 - 112 MB | NPU | -- |
| YOLOv8-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 4.906 ms | 0 - 37 MB | NPU | -- |
| YOLOv8-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 4.338 ms | 5 - 27 MB | NPU | -- |
| YOLOv8-Segmentation | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 6.657 ms | 3 - 104 MB | NPU | -- |
| YOLOv8-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 3.644 ms | 0 - 96 MB | NPU | -- |
| YOLOv8-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 3.224 ms | 5 - 204 MB | NPU | -- |
| YOLOv8-Segmentation | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 4.32 ms | 0 - 190 MB | NPU | -- |
| YOLOv8-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 3.475 ms | 3 - 76 MB | NPU | -- |
| YOLOv8-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 2.858 ms | 5 - 129 MB | NPU | -- |
| YOLOv8-Segmentation | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 3.545 ms | 14 - 109 MB | NPU | -- |
| YOLOv8-Segmentation | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 4.749 ms | 93 - 93 MB | NPU | -- |
| YOLOv8-Segmentation | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 6.455 ms | 17 - 17 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 7.783 ms | 2 - 33 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 4.76 ms | 2 - 47 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 3.791 ms | 2 - 13 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 47.221 ms | 19 - 172 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 4.412 ms | 2 - 33 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 16.316 ms | 2 - 38 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX | 220.261 ms | 87 - 108 MB | CPU | -- |
| YOLOv8-Segmentation | w8a16 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 187.916 ms | 91 - 97 MB | CPU | -- |
| YOLOv8-Segmentation | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 7.783 ms | 2 - 33 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 3.801 ms | 2 - 13 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 5.147 ms | 2 - 42 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 3.82 ms | 2 - 12 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 4.412 ms | 2 - 33 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 3.812 ms | 2 - 13 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 48.642 ms | 27 - 174 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 2.534 ms | 2 - 43 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 37.366 ms | 19 - 1509 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 2.102 ms | 2 - 41 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 40.583 ms | 17 - 553 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 4.239 ms | 7 - 7 MB | NPU | -- |
| YOLOv8-Segmentation | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 47.26 ms | 59 - 59 MB | NPU | -- |
Installation
Install the package via pip:
pip install "qai-hub-models[yolov8-seg]"
Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
Sign-in to Qualcomm® AI Hub with your
Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.
With this API token, you can configure your client to run models on the cloud hosted devices.
qai-hub configure --api_token API_TOKEN
Navigate to docs for more information.
Demo off target
The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.
python -m qai_hub_models.models.yolov8_seg.demo
The above demo runs a reference implementation of pre-processing, model inference, and post processing.
NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).
%run -m qai_hub_models.models.yolov8_seg.demo
Run model on a cloud-hosted device
In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:
- Performance check on-device on a cloud-hosted device
- Downloads compiled assets that can be deployed on-device for Android.
- Accuracy check between PyTorch and on-device outputs.
python -m qai_hub_models.models.yolov8_seg.export
How does this work?
This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:
Step 1: Compile model for on-device deployment
To compile a PyTorch model for on-device deployment, we first trace the model
in memory using the jit.trace and then call the submit_compile_job API.
import torch
import qai_hub as hub
from qai_hub_models.models.yolov8_seg import Model
# Load the model
torch_model = Model.from_pretrained()
# Device
device = hub.Device("Samsung Galaxy S24")
# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()
pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
# Compile model on a specific device
compile_job = hub.submit_compile_job(
model=pt_model,
device=device,
input_specs=torch_model.get_input_spec(),
)
# Get target model to run on-device
target_model = compile_job.get_target_model()
Step 2: Performance profiling on cloud-hosted device
After compiling models from step 1. Models can be profiled model on-device using the
target_model. Note that this scripts runs the model on a device automatically
provisioned in the cloud. Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
profile_job = hub.submit_profile_job(
model=target_model,
device=device,
)
Step 3: Verify on-device accuracy
To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.
input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
model=target_model,
device=device,
inputs=input_data,
)
on_device_output = inference_job.download_output_data()
With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.
Note: This on-device profiling and inference requires access to Qualcomm® AI Hub. Sign up for access.
Run demo on a cloud-hosted device
You can also run the demo on-device.
python -m qai_hub_models.models.yolov8_seg.demo --eval-mode on-device
NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).
%run -m qai_hub_models.models.yolov8_seg.demo -- --eval-mode on-device
Deploying compiled model to Android
The models can be deployed using multiple runtimes:
TensorFlow Lite (
.tfliteexport): This tutorial provides a guide to deploy the .tflite model in an Android application.QNN (
.soexport ): This sample app provides instructions on how to use the.soshared library in an Android application.
View on Qualcomm® AI Hub
Get more details on YOLOv8-Segmentation's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of YOLOv8-Segmentation can be found here.
- The license for the compiled assets for on-device deployment 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.