qualcomm
/

Yolo-v5

@@ -11,258 +11,88 @@ pipeline_tag: object-detection
 ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolov5/web-assets/model_demo.png)
-# Yolo-v5: Optimized for Mobile Deployment
-## Real-time object detection optimized for mobile and edge
 YoloV5 is a machine learning model that predicts bounding boxes and classes of objects in an image.
-This model is an implementation of Yolo-v5 found [here](https://github.com/ultralytics/yolov5).
-This repository provides scripts to run Yolo-v5 on Qualcomm® devices.
-More details on model performance across various devices, can be found
-[here](https://aihub.qualcomm.com/models/yolov5).
-**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: YoloV5-M
-  - Input resolution: 640x640
-  - Number of parameters: 21.2M
-  - Model size (float): 81.1 MB
-  - Model size (w8a16): 21.8 MB
-| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
-|---|---|---|---|---|---|---|---|---|
-| Yolo-v5 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 63.145 ms | 0 - 230 MB | NPU | -- |
-| Yolo-v5 | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 63.271 ms | 4 - 206 MB | NPU | -- |
-| Yolo-v5 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 25.367 ms | 0 - 287 MB | NPU | -- |
-| Yolo-v5 | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 25.75 ms | 5 - 238 MB | NPU | -- |
-| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 11.045 ms | 0 - 3 MB | NPU | -- |
-| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 10.901 ms | 5 - 7 MB | NPU | -- |
-| Yolo-v5 | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 13.099 ms | 0 - 54 MB | NPU | -- |
-| Yolo-v5 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 77.466 ms | 0 - 214 MB | NPU | -- |
-| Yolo-v5 | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 17.082 ms | 1 - 201 MB | NPU | -- |
-| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 8.225 ms | 0 - 393 MB | NPU | -- |
-| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 8.27 ms | 5 - 315 MB | NPU | -- |
-| Yolo-v5 | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 9.653 ms | 5 - 213 MB | NPU | -- |
-| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 6.181 ms | 0 - 209 MB | NPU | -- |
-| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 6.172 ms | 5 - 209 MB | NPU | -- |
-| Yolo-v5 | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 7.41 ms | 1 - 167 MB | NPU | -- |
-| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 4.473 ms | 0 - 214 MB | NPU | -- |
-| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 4.515 ms | 5 - 224 MB | NPU | -- |
-| Yolo-v5 | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 5.915 ms | 5 - 162 MB | NPU | -- |
-| Yolo-v5 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 11.534 ms | 5 - 5 MB | NPU | -- |
-| Yolo-v5 | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 13.346 ms | 46 - 46 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | QNN_DLC | 122.314 ms | 2 - 244 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | ONNX | 837.633 ms | 121 - 138 MB | CPU | -- |
-| Yolo-v5 | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 31.692 ms | 2 - 6 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 1723.454 ms | 91 - 99 MB | CPU | -- |
-| Yolo-v5 | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 20.943 ms | 2 - 206 MB | NPU | -- |
-| Yolo-v5 | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 12.011 ms | 2 - 260 MB | NPU | -- |
-| Yolo-v5 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 8.646 ms | 2 - 5 MB | NPU | -- |
-| Yolo-v5 | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 11.495 ms | 0 - 29 MB | NPU | -- |
-| Yolo-v5 | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 9.017 ms | 1 - 205 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 5.785 ms | 2 - 260 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 7.703 ms | 3 - 254 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 4.357 ms | 2 - 209 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 5.075 ms | 0 - 203 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 12.531 ms | 2 - 225 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 817.677 ms | 161 - 178 MB | CPU | -- |
-| Yolo-v5 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 3.296 ms | 2 - 218 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 4.356 ms | 3 - 209 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 9.357 ms | 2 - 2 MB | NPU | -- |
-| Yolo-v5 | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 11.509 ms | 23 - 23 MB | NPU | -- |
-## Installation
-Install the package via pip:
-```bash
-# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
-pip install "qai-hub-models[yolov5]"
-```
-## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
-Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) 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.
-```bash
-qai-hub configure --api_token API_TOKEN
-```
-Navigate to [docs](https://workbench.aihub.qualcomm.com/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.
-```bash
-python -m qai_hub_models.models.yolov5.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.yolov5.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.
-```bash
-python -m qai_hub_models.models.yolov5.export
-```
-## How does this work?
-This [export script](https://aihub.qualcomm.com/models/yolov5/qai_hub_models/models/Yolo-v5/export.py)
-leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) 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.
-```python
-import torch
-import qai_hub as hub
-from qai_hub_models.models.yolov5 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.
-```python
-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.
-```python
-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](https://myaccount.qualcomm.com/signup).
-## Run demo on a cloud-hosted device
-You can also run the demo on-device.
-```bash
-python -m qai_hub_models.models.yolov5.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.yolov5.demo -- --eval-mode on-device
-```
-## Deploying compiled model to Android
-The models can be deployed using multiple runtimes:
-- TensorFlow Lite (`.tflite` export): [This
-  tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
-  guide to deploy the .tflite model in an Android application.
-- QNN (`.so` export ): This [sample
-  app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
-provides instructions on how to use the `.so` shared library  in an Android application.
-## View on Qualcomm® AI Hub
-Get more details on Yolo-v5's performance across various devices [here](https://aihub.qualcomm.com/models/yolov5).
-Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
 ## License
 * The license for the original implementation of Yolo-v5 can be found
   [here](https://github.com/ultralytics/yolov5?tab=AGPL-3.0-1-ov-file#readme).
 ## References
 * [Source Model Implementation](https://github.com/ultralytics/yolov5)
 ## Community
 * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
 * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).

 ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolov5/web-assets/model_demo.png)
+# Yolo-v5: Optimized for Qualcomm Devices
 YoloV5 is a machine learning model that predicts bounding boxes and classes of objects in an image.
+This is based on the implementation of Yolo-v5 found [here](https://github.com/ultralytics/yolov5).
+This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov5) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
+Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) 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](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov5) 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 [Yolo-v5 on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov5) for usage instructions.
+## Model Details
+**Model Type:** Model_use_case.object_detection
+**Model Stats:**
+- Model checkpoint: YoloV5-M
+- Input resolution: 640x640
+- Number of parameters: 21.2M
+- Model size (float): 81.1 MB
+- Model size (w8a16): 21.8 MB
+## Performance Summary
+| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
+|---|---|---|---|---|---|---
+| Yolo-v5 | ONNX | float | Snapdragon® X Elite | 13.392 ms | 46 - 46 MB | NPU
+| Yolo-v5 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 9.711 ms | 1 - 218 MB | NPU
+| Yolo-v5 | ONNX | float | Qualcomm® QCS8550 (Proxy) | 12.922 ms | 0 - 54 MB | NPU
+| Yolo-v5 | ONNX | float | Qualcomm® QCS9075 | 21.799 ms | 5 - 12 MB | NPU
+| Yolo-v5 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 7.358 ms | 0 - 163 MB | NPU
+| Yolo-v5 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.934 ms | 2 - 156 MB | NPU
+| Yolo-v5 | ONNX | w8a16 | Snapdragon® X Elite | 11.535 ms | 24 - 24 MB | NPU
+| Yolo-v5 | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 7.7 ms | 3 - 253 MB | NPU
+| Yolo-v5 | ONNX | w8a16 | Qualcomm® QCS6490 | 1734.938 ms | 93 - 103 MB | CPU
+| Yolo-v5 | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 11.586 ms | 0 - 28 MB | NPU
+| Yolo-v5 | ONNX | w8a16 | Qualcomm® QCS9075 | 12.749 ms | 2 - 5 MB | NPU
+| Yolo-v5 | ONNX | w8a16 | Qualcomm® QCM6690 | 844.595 ms | 97 - 106 MB | CPU
+| Yolo-v5 | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 5.083 ms | 0 - 201 MB | NPU
+| Yolo-v5 | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 817.456 ms | 97 - 108 MB | CPU
+| Yolo-v5 | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.928 ms | 3 - 209 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Snapdragon® X Elite | 11.954 ms | 5 - 5 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 8.492 ms | 0 - 273 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 63.73 ms | 1 - 214 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 11.34 ms | 5 - 7 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Qualcomm® QCS9075 | 18.467 ms | 5 - 11 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 26.11 ms | 5 - 263 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.344 ms | 0 - 214 MB | NPU
+| Yolo-v5 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.773 ms | 5 - 214 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Snapdragon® X Elite | 9.458 ms | 2 - 2 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 5.885 ms | 2 - 285 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 32.048 ms | 1 - 5 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 20.959 ms | 1 - 232 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 8.753 ms | 2 - 4 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 10.908 ms | 2 - 6 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 124.027 ms | 2 - 269 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 13.553 ms | 2 - 285 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 4.192 ms | 2 - 239 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 12.722 ms | 2 - 251 MB | NPU
+| Yolo-v5 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.234 ms | 2 - 246 MB | NPU
+| Yolo-v5 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 8.197 ms | 0 - 315 MB | NPU
+| Yolo-v5 | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 63.071 ms | 0 - 236 MB | NPU
+| Yolo-v5 | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 10.819 ms | 0 - 3 MB | NPU
+| Yolo-v5 | TFLITE | float | Qualcomm® QCS9075 | 17.898 ms | 0 - 57 MB | NPU
+| Yolo-v5 | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 30.761 ms | 0 - 314 MB | NPU
+| Yolo-v5 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.127 ms | 0 - 240 MB | NPU
+| Yolo-v5 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.317 ms | 0 - 238 MB | NPU
 ## License
 * The license for the original implementation of Yolo-v5 can be found
   [here](https://github.com/ultralytics/yolov5?tab=AGPL-3.0-1-ov-file#readme).
 ## References
 * [Source Model Implementation](https://github.com/ultralytics/yolov5)
 ## Community
 * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
 * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).