Yolo-v7: Optimized for Mobile Deployment
Real-time object detection optimized for mobile and edge
YoloV7 is a machine learning model that predicts bounding boxes and classes of objects in an image.
This model is an implementation of Yolo-v7 found here.
This repository provides scripts to run Yolo-v7 on Qualcomm® devices. More details on model performance across various devices, can be found here.
Model Details
- Model Type: Object detection
- Model Stats:
- Model checkpoint: YoloV7 Tiny
- Input resolution: 720p (720x1280)
- Number of parameters: 6.39M
- Model size: 24.4 MB
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| Yolo-v7 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 17.203 ms | 0 - 10 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 10.583 ms | 5 - 21 MB | FP16 | NPU | Yolo-v7.so |
| Yolo-v7 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 12.221 ms | 0 - 13 MB | FP16 | NPU | Yolo-v7.onnx |
| Yolo-v7 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 11.527 ms | 0 - 99 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 7.099 ms | 5 - 75 MB | FP16 | NPU | Yolo-v7.so |
| Yolo-v7 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 7.984 ms | 6 - 124 MB | FP16 | NPU | Yolo-v7.onnx |
| Yolo-v7 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 12.198 ms | 1 - 70 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 7.058 ms | 5 - 69 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 8.162 ms | 2 - 81 MB | FP16 | NPU | Yolo-v7.onnx |
| Yolo-v7 | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 17.066 ms | 1 - 4 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 10.342 ms | 5 - 6 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 17.232 ms | 0 - 2 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | SA8255 (Proxy) | SA8255P Proxy | QNN | 10.441 ms | 5 - 6 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 17.182 ms | 1 - 3 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | SA8775 (Proxy) | SA8775P Proxy | QNN | 10.337 ms | 5 - 6 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 17.184 ms | 1 - 3 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | SA8650 (Proxy) | SA8650P Proxy | QNN | 10.343 ms | 5 - 6 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | SA8295P ADP | SA8295P | TFLITE | 21.595 ms | 1 - 73 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | SA8295P ADP | SA8295P | QNN | 13.697 ms | 0 - 5 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 19.511 ms | 1 - 103 MB | FP16 | NPU | Yolo-v7.tflite |
| Yolo-v7 | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 12.655 ms | 5 - 62 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 10.935 ms | 5 - 5 MB | FP16 | NPU | Use Export Script |
| Yolo-v7 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 14.1 ms | 9 - 9 MB | FP16 | NPU | Yolo-v7.onnx |
Installation
This model can be installed as a Python package via pip.
pip install "qai-hub-models[yolov7]"
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.yolov7.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.yolov7.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.yolov7.export
Profiling Results
------------------------------------------------------------
Yolo-v7
Device : Samsung Galaxy S23 (13)
Runtime : TFLITE
Estimated inference time (ms) : 17.2
Estimated peak memory usage (MB): [0, 10]
Total # Ops : 215
Compute Unit(s) : NPU (215 ops)
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.yolov7 import
# Load the model
# Device
device = hub.Device("Samsung Galaxy S23")
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.yolov7.demo --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.yolov7.demo -- --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 Yolo-v7's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of Yolo-v7 can be found here.
- The license for the compiled assets for on-device deployment can be found here
References
- YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
- Source Model Implementation
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.