YOLOv10: Real-Time End-to-End Object Detection
Paper
•
2405.14458
•
Published
•
6
YOLOv10-Detection: Optimized for Mobile Deployment
Real-time object detection optimized for mobile and edge by Ultralytics
Ultralytics YOLOv10 is a machine learning model that predicts bounding boxes and classes of objects in an image.
This model is an implementation of YOLOv10-Detection found here.
This repository provides scripts to run YOLOv10-Detection 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.object_detection
- Model Stats:
- Model checkpoint: YOLOv10-N
- Input resolution: 640x640
- Number of parameters: 2.33M
- Model size (float): 8.95 MB
- Model size (w8a8): 2.55 MB
- Model size (w8a16): 3.04 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| YOLOv10-Detection | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 12.71 ms | 0 - 224 MB | NPU | -- |
| YOLOv10-Detection | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 12.644 ms | 4 - 209 MB | NPU | -- |
| YOLOv10-Detection | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 7.705 ms | 0 - 184 MB | NPU | -- |
| YOLOv10-Detection | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 7.729 ms | 5 - 183 MB | NPU | -- |
| YOLOv10-Detection | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 3.702 ms | 0 - 3 MB | NPU | -- |
| YOLOv10-Detection | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 3.635 ms | 5 - 8 MB | NPU | -- |
| YOLOv10-Detection | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 5.819 ms | 5 - 10 MB | NPU | -- |
| YOLOv10-Detection | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 5.21 ms | 0 - 203 MB | NPU | -- |
| YOLOv10-Detection | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 5.186 ms | 1 - 210 MB | NPU | -- |
| YOLOv10-Detection | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 12.71 ms | 0 - 224 MB | NPU | -- |
| YOLOv10-Detection | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 12.644 ms | 4 - 209 MB | NPU | -- |
| YOLOv10-Detection | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 3.715 ms | 0 - 3 MB | NPU | -- |
| YOLOv10-Detection | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 3.656 ms | 5 - 8 MB | NPU | -- |
| YOLOv10-Detection | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 8.178 ms | 0 - 160 MB | NPU | -- |
| YOLOv10-Detection | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 8.066 ms | 0 - 154 MB | NPU | -- |
| YOLOv10-Detection | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 3.707 ms | 0 - 2 MB | NPU | -- |
| YOLOv10-Detection | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 3.639 ms | 4 - 6 MB | NPU | -- |
| YOLOv10-Detection | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 5.21 ms | 0 - 203 MB | NPU | -- |
| YOLOv10-Detection | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 5.186 ms | 1 - 210 MB | NPU | -- |
| YOLOv10-Detection | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 2.674 ms | 0 - 395 MB | NPU | -- |
| YOLOv10-Detection | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 2.68 ms | 4 - 362 MB | NPU | -- |
| YOLOv10-Detection | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 3.708 ms | 0 - 215 MB | NPU | -- |
| YOLOv10-Detection | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 2.105 ms | 0 - 211 MB | NPU | -- |
| YOLOv10-Detection | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 2.049 ms | 5 - 214 MB | NPU | -- |
| YOLOv10-Detection | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 3.509 ms | 1 - 176 MB | NPU | -- |
| YOLOv10-Detection | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 1.61 ms | 0 - 223 MB | NPU | -- |
| YOLOv10-Detection | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 1.61 ms | 5 - 233 MB | NPU | -- |
| YOLOv10-Detection | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 2.558 ms | 2 - 150 MB | NPU | -- |
| YOLOv10-Detection | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 4.089 ms | 5 - 5 MB | NPU | -- |
| YOLOv10-Detection | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.999 ms | 5 - 5 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | QNN_DLC | 21.463 ms | 2 - 159 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | ONNX | 177.989 ms | 67 - 82 MB | CPU | -- |
| YOLOv10-Detection | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 350.649 ms | 67 - 73 MB | CPU | -- |
| YOLOv10-Detection | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 8.02 ms | 2 - 153 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 4.305 ms | 2 - 5 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 5.345 ms | 2 - 7 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 4.914 ms | 1 - 152 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 140.17 ms | 63 - 67 MB | CPU | -- |
| YOLOv10-Detection | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 8.02 ms | 2 - 153 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 4.308 ms | 2 - 4 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 4.315 ms | 1 - 3 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 4.914 ms | 1 - 152 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 3.061 ms | 2 - 180 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 3.317 ms | 0 - 165 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 2.157 ms | 2 - 160 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 2.549 ms | 0 - 136 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 4.437 ms | 2 - 161 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 158.0 ms | 69 - 86 MB | CPU | -- |
| YOLOv10-Detection | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 1.882 ms | 2 - 159 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 2.305 ms | 1 - 145 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 4.654 ms | 2 - 2 MB | NPU | -- |
| YOLOv10-Detection | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.482 ms | 2 - 2 MB | NPU | -- |
Installation
Install the package via pip:
# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install "qai-hub-models[yolov10-det]"
Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
Sign-in to Qualcomm® AI Hub Workbench 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.yolov10_det.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.yolov10_det.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.yolov10_det.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.yolov10_det import Model
# Load the model
torch_model = Model.from_pretrained()
# Device
device = hub.Device("Samsung Galaxy S25")
# 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 Workbench. 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.yolov10_det.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.yolov10_det.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 YOLOv10-Detection's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of YOLOv10-Detection 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.