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.

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: YoloV7 Tiny
- Input resolution: 640x640
- Number of parameters: 6.24M
- Model size (float): 23.8 MB
- Model size (w8a8): 6.23 MB
- Model size (w8a16): 6.66 MB

Model	Precision	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit	Target Model
Yolo-v7	float	QCS8275 (Proxy)	Qualcomm® QCS8275 (Proxy)	TFLITE	29.023 ms	1 - 135 MB	NPU	--
Yolo-v7	float	QCS8275 (Proxy)	Qualcomm® QCS8275 (Proxy)	QNN_DLC	18.738 ms	0 - 267 MB	NPU	--
Yolo-v7	float	QCS8450 (Proxy)	Qualcomm® QCS8450 (Proxy)	TFLITE	22.496 ms	1 - 173 MB	NPU	--
Yolo-v7	float	QCS8450 (Proxy)	Qualcomm® QCS8450 (Proxy)	QNN_DLC	10.64 ms	5 - 165 MB	NPU	--
Yolo-v7	float	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	TFLITE	13.977 ms	1 - 3 MB	NPU	--
Yolo-v7	float	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	QNN_DLC	6.464 ms	5 - 7 MB	NPU	--
Yolo-v7	float	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	ONNX	8.321 ms	0 - 12 MB	NPU	--
Yolo-v7	float	QCS9075 (Proxy)	Qualcomm® QCS9075 (Proxy)	TFLITE	15.564 ms	1 - 137 MB	NPU	--
Yolo-v7	float	QCS9075 (Proxy)	Qualcomm® QCS9075 (Proxy)	QNN_DLC	8.118 ms	1 - 257 MB	NPU	--
Yolo-v7	float	SA7255P ADP	Qualcomm® SA7255P	TFLITE	29.023 ms	1 - 135 MB	NPU	--
Yolo-v7	float	SA7255P ADP	Qualcomm® SA7255P	QNN_DLC	18.738 ms	0 - 267 MB	NPU	--
Yolo-v7	float	SA8255 (Proxy)	Qualcomm® SA8255P (Proxy)	TFLITE	14.018 ms	1 - 3 MB	NPU	--
Yolo-v7	float	SA8255 (Proxy)	Qualcomm® SA8255P (Proxy)	QNN_DLC	6.453 ms	5 - 7 MB	NPU	--
Yolo-v7	float	SA8295P ADP	Qualcomm® SA8295P	TFLITE	18.452 ms	1 - 145 MB	NPU	--
Yolo-v7	float	SA8295P ADP	Qualcomm® SA8295P	QNN_DLC	9.217 ms	0 - 147 MB	NPU	--
Yolo-v7	float	SA8650 (Proxy)	Qualcomm® SA8650P (Proxy)	TFLITE	14.329 ms	1 - 3 MB	NPU	--
Yolo-v7	float	SA8650 (Proxy)	Qualcomm® SA8650P (Proxy)	QNN_DLC	6.448 ms	5 - 7 MB	NPU	--
Yolo-v7	float	SA8775P ADP	Qualcomm® SA8775P	TFLITE	15.564 ms	1 - 137 MB	NPU	--
Yolo-v7	float	SA8775P ADP	Qualcomm® SA8775P	QNN_DLC	8.118 ms	1 - 257 MB	NPU	--
Yolo-v7	float	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	TFLITE	10.814 ms	0 - 165 MB	NPU	--
Yolo-v7	float	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	QNN_DLC	4.549 ms	5 - 491 MB	NPU	--
Yolo-v7	float	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	ONNX	5.621 ms	6 - 461 MB	NPU	--
Yolo-v7	float	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	TFLITE	16.649 ms	0 - 135 MB	GPU	--
Yolo-v7	float	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	QNN_DLC	3.539 ms	0 - 237 MB	NPU	--
Yolo-v7	float	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	ONNX	5.363 ms	3 - 224 MB	NPU	--
Yolo-v7	float	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	TFLITE	6.522 ms	1 - 145 MB	NPU	--
Yolo-v7	float	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	QNN_DLC	2.875 ms	5 - 263 MB	NPU	--
Yolo-v7	float	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	ONNX	3.64 ms	0 - 224 MB	NPU	--
Yolo-v7	float	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN_DLC	6.891 ms	5 - 5 MB	NPU	--
Yolo-v7	float	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	8.47 ms	9 - 9 MB	NPU	--
Yolo-v7	w8a16	Dragonwing Q-6690 MTP	Qualcomm® Qcm6690	QNN_DLC	31.878 ms	2 - 250 MB	NPU	--
Yolo-v7	w8a16	Dragonwing Q-6690 MTP	Qualcomm® Qcm6690	ONNX	215.545 ms	83 - 97 MB	CPU	--
Yolo-v7	w8a16	Dragonwing RB3 Gen 2 Vision Kit	Qualcomm® QCS6490	QNN_DLC	11.435 ms	2 - 6 MB	NPU	--
Yolo-v7	w8a16	Dragonwing RB3 Gen 2 Vision Kit	Qualcomm® QCS6490	ONNX	437.653 ms	83 - 88 MB	CPU	--
Yolo-v7	w8a16	QCS8275 (Proxy)	Qualcomm® QCS8275 (Proxy)	QNN_DLC	8.991 ms	2 - 299 MB	NPU	--
Yolo-v7	w8a16	QCS8450 (Proxy)	Qualcomm® QCS8450 (Proxy)	QNN_DLC	6.094 ms	2 - 331 MB	NPU	--
Yolo-v7	w8a16	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	QNN_DLC	3.927 ms	2 - 5 MB	NPU	--
Yolo-v7	w8a16	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	ONNX	5.196 ms	0 - 7 MB	NPU	--
Yolo-v7	w8a16	QCS9075 (Proxy)	Qualcomm® QCS9075 (Proxy)	QNN_DLC	4.563 ms	0 - 301 MB	NPU	--
Yolo-v7	w8a16	RB5 (Proxy)	Qualcomm® QCS8250 (Proxy)	ONNX	180.457 ms	81 - 89 MB	CPU	--
Yolo-v7	w8a16	SA7255P ADP	Qualcomm® SA7255P	QNN_DLC	8.991 ms	2 - 299 MB	NPU	--
Yolo-v7	w8a16	SA8255 (Proxy)	Qualcomm® SA8255P (Proxy)	QNN_DLC	3.937 ms	2 - 5 MB	NPU	--
Yolo-v7	w8a16	SA8295P ADP	Qualcomm® SA8295P	QNN_DLC	5.677 ms	0 - 299 MB	NPU	--
Yolo-v7	w8a16	SA8650 (Proxy)	Qualcomm® SA8650P (Proxy)	QNN_DLC	3.928 ms	3 - 5 MB	NPU	--
Yolo-v7	w8a16	SA8775P ADP	Qualcomm® SA8775P	QNN_DLC	4.563 ms	0 - 301 MB	NPU	--
Yolo-v7	w8a16	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	QNN_DLC	2.818 ms	2 - 328 MB	NPU	--
Yolo-v7	w8a16	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	ONNX	3.481 ms	3 - 327 MB	NPU	--
Yolo-v7	w8a16	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	QNN_DLC	2.061 ms	2 - 304 MB	NPU	--
Yolo-v7	w8a16	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	ONNX	2.67 ms	0 - 283 MB	NPU	--
Yolo-v7	w8a16	Snapdragon 7 Gen 4 QRD	Snapdragon® 7 Gen 4 Mobile	QNN_DLC	5.027 ms	2 - 233 MB	NPU	--
Yolo-v7	w8a16	Snapdragon 7 Gen 4 QRD	Snapdragon® 7 Gen 4 Mobile	ONNX	201.854 ms	82 - 99 MB	CPU	--
Yolo-v7	w8a16	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	QNN_DLC	1.613 ms	2 - 298 MB	NPU	--
Yolo-v7	w8a16	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	ONNX	2.143 ms	0 - 283 MB	NPU	--
Yolo-v7	w8a16	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN_DLC	4.432 ms	2 - 2 MB	NPU	--
Yolo-v7	w8a16	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	5.419 ms	5 - 5 MB	NPU	--
Yolo-v7	w8a8	Dragonwing Q-6690 MTP	Qualcomm® Qcm6690	TFLITE	14.407 ms	0 - 148 MB	NPU	--
Yolo-v7	w8a8	Dragonwing Q-6690 MTP	Qualcomm® Qcm6690	QNN_DLC	14.621 ms	1 - 167 MB	NPU	--
Yolo-v7	w8a8	Dragonwing Q-6690 MTP	Qualcomm® Qcm6690	ONNX	54.455 ms	39 - 55 MB	CPU	--
Yolo-v7	w8a8	Dragonwing RB3 Gen 2 Vision Kit	Qualcomm® QCS6490	TFLITE	6.113 ms	0 - 9 MB	NPU	--
Yolo-v7	w8a8	Dragonwing RB3 Gen 2 Vision Kit	Qualcomm® QCS6490	QNN_DLC	6.251 ms	3 - 6 MB	NPU	--
Yolo-v7	w8a8	Dragonwing RB3 Gen 2 Vision Kit	Qualcomm® QCS6490	ONNX	74.368 ms	38 - 45 MB	CPU	--
Yolo-v7	w8a8	QCS8275 (Proxy)	Qualcomm® QCS8275 (Proxy)	TFLITE	4.74 ms	0 - 154 MB	NPU	--
Yolo-v7	w8a8	QCS8275 (Proxy)	Qualcomm® QCS8275 (Proxy)	QNN_DLC	4.592 ms	1 - 147 MB	NPU	--
Yolo-v7	w8a8	QCS8450 (Proxy)	Qualcomm® QCS8450 (Proxy)	TFLITE	3.118 ms	0 - 162 MB	NPU	--
Yolo-v7	w8a8	QCS8450 (Proxy)	Qualcomm® QCS8450 (Proxy)	QNN_DLC	2.863 ms	1 - 174 MB	NPU	--
Yolo-v7	w8a8	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	TFLITE	2.237 ms	0 - 2 MB	NPU	--
Yolo-v7	w8a8	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	QNN_DLC	2.139 ms	1 - 4 MB	NPU	--
Yolo-v7	w8a8	QCS8550 (Proxy)	Qualcomm® QCS8550 (Proxy)	ONNX	2.708 ms	0 - 8 MB	NPU	--
Yolo-v7	w8a8	QCS9075 (Proxy)	Qualcomm® QCS9075 (Proxy)	TFLITE	2.67 ms	0 - 156 MB	NPU	--
Yolo-v7	w8a8	QCS9075 (Proxy)	Qualcomm® QCS9075 (Proxy)	QNN_DLC	2.523 ms	1 - 148 MB	NPU	--
Yolo-v7	w8a8	RB5 (Proxy)	Qualcomm® QCS8250 (Proxy)	TFLITE	57.504 ms	15 - 53 MB	GPU	--
Yolo-v7	w8a8	RB5 (Proxy)	Qualcomm® QCS8250 (Proxy)	ONNX	50.459 ms	34 - 43 MB	CPU	--
Yolo-v7	w8a8	SA7255P ADP	Qualcomm® SA7255P	TFLITE	4.74 ms	0 - 154 MB	NPU	--
Yolo-v7	w8a8	SA7255P ADP	Qualcomm® SA7255P	QNN_DLC	4.592 ms	1 - 147 MB	NPU	--
Yolo-v7	w8a8	SA8255 (Proxy)	Qualcomm® SA8255P (Proxy)	TFLITE	2.234 ms	0 - 2 MB	NPU	--
Yolo-v7	w8a8	SA8255 (Proxy)	Qualcomm® SA8255P (Proxy)	QNN_DLC	2.147 ms	0 - 2 MB	NPU	--
Yolo-v7	w8a8	SA8295P ADP	Qualcomm® SA8295P	TFLITE	3.703 ms	0 - 146 MB	NPU	--
Yolo-v7	w8a8	SA8295P ADP	Qualcomm® SA8295P	QNN_DLC	3.5 ms	0 - 155 MB	NPU	--
Yolo-v7	w8a8	SA8650 (Proxy)	Qualcomm® SA8650P (Proxy)	TFLITE	2.237 ms	0 - 2 MB	NPU	--
Yolo-v7	w8a8	SA8650 (Proxy)	Qualcomm® SA8650P (Proxy)	QNN_DLC	2.142 ms	1 - 3 MB	NPU	--
Yolo-v7	w8a8	SA8775P ADP	Qualcomm® SA8775P	TFLITE	2.67 ms	0 - 156 MB	NPU	--
Yolo-v7	w8a8	SA8775P ADP	Qualcomm® SA8775P	QNN_DLC	2.523 ms	1 - 148 MB	NPU	--
Yolo-v7	w8a8	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	TFLITE	1.424 ms	0 - 178 MB	NPU	--
Yolo-v7	w8a8	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	QNN_DLC	1.417 ms	1 - 171 MB	NPU	--
Yolo-v7	w8a8	Samsung Galaxy S24	Snapdragon® 8 Gen 3 Mobile	ONNX	1.744 ms	0 - 159 MB	NPU	--
Yolo-v7	w8a8	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	TFLITE	1.214 ms	0 - 161 MB	NPU	--
Yolo-v7	w8a8	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	QNN_DLC	1.053 ms	1 - 152 MB	NPU	--
Yolo-v7	w8a8	Samsung Galaxy S25	Snapdragon® 8 Elite For Galaxy Mobile	ONNX	1.391 ms	0 - 135 MB	NPU	--
Yolo-v7	w8a8	Snapdragon 7 Gen 4 QRD	Snapdragon® 7 Gen 4 Mobile	TFLITE	2.636 ms	0 - 148 MB	NPU	--
Yolo-v7	w8a8	Snapdragon 7 Gen 4 QRD	Snapdragon® 7 Gen 4 Mobile	QNN_DLC	2.643 ms	1 - 296 MB	NPU	--
Yolo-v7	w8a8	Snapdragon 7 Gen 4 QRD	Snapdragon® 7 Gen 4 Mobile	ONNX	50.926 ms	41 - 58 MB	CPU	--
Yolo-v7	w8a8	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	TFLITE	0.982 ms	0 - 176 MB	NPU	--
Yolo-v7	w8a8	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	QNN_DLC	0.895 ms	1 - 151 MB	NPU	--
Yolo-v7	w8a8	Snapdragon 8 Elite Gen 5 QRD	Snapdragon® 8 Elite Gen5 Mobile	ONNX	1.245 ms	0 - 135 MB	NPU	--
Yolo-v7	w8a8	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN_DLC	2.397 ms	1 - 1 MB	NPU	--
Yolo-v7	w8a8	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	2.752 ms	5 - 5 MB	NPU	--

Installation

Install the package via pip:

# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install "qai-hub-models[yolov7]"

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.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

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 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.yolov7.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.yolov7.demo -- --eval-mode on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.
QNN (.so export ): This sample app provides instructions on how to use the .so shared 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.

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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/Yolo-v7

YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

Paper • 2207.02696 • Published Jul 6, 2022