---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: object-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/web-assets/model_demo.png)

# Yolo-X: Optimized for Qualcomm Devices

YoloX is a machine learning model that predicts bounding boxes and classes of objects in an image.

This is based on the implementation of Yolo-X found [here](https://github.com/Megvii-BaseDetection/YOLOX/).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/yolox) 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
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-onnx-w8a16.zip)
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-onnx-w8a8.zip)
| ONNX | w8a8_mixed_int16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-onnx-w8a8_mixed_int16.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-qnn_dlc-w8a16.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-qnn_dlc-w8a8.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yolox/releases/v0.59.0/yolox-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[Yolo-X on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/yolox)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/yolox) 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

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [Yolo-X on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/yolox) for usage instructions.

## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- Model checkpoint: YoloX Small
- Input resolution: 640x640
- Number of parameters: 8.98M
- Model size (float): 34.3 MB
- Model size (w8a16): 9.53 MB
- Model size (w8a8): 8.96 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| Yolo-X | ONNX | float | Snapdragon® X2 Elite | 7.712 ms | 5 - 5 MB | NPU
| Yolo-X | ONNX | float | Snapdragon® X Elite | 13.489 ms | 14 - 14 MB | NPU
| Yolo-X | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 9.265 ms | 5 - 187 MB | NPU
| Yolo-X | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 25.641 ms | 5 - 213 MB | NPU
| Yolo-X | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 13.702 ms | 0 - 158 MB | NPU
| Yolo-X | ONNX | float | Qualcomm® QCS8450 | 25.641 ms | 5 - 213 MB | NPU
| Yolo-X | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 16.891 ms | 5 - 7 MB | NPU
| Yolo-X | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 13.489 ms | 14 - 14 MB | NPU
| Yolo-X | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 7.727 ms | 1 - 159 MB | NPU
| Yolo-X | ONNX | float | Snapdragon® 8 Elite Mobile | 7.727 ms | 1 - 159 MB | NPU
| Yolo-X | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.276 ms | 0 - 160 MB | NPU
| Yolo-X | ONNX | w8a16 | Snapdragon® X2 Elite | 11.412 ms | 2 - 2 MB | NPU
| Yolo-X | ONNX | w8a16 | Snapdragon® X Elite | 14.985 ms | 7 - 7 MB | NPU
| Yolo-X | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 10.055 ms | 3 - 239 MB | NPU
| Yolo-X | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 16.195 ms | 3 - 240 MB | NPU
| Yolo-X | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 29.397 ms | 2 - 5 MB | NPU
| Yolo-X | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 14.24 ms | 0 - 85 MB | NPU
| Yolo-X | ONNX | w8a16 | Qualcomm® QCS8450 | 16.195 ms | 3 - 240 MB | NPU
| Yolo-X | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 15.441 ms | 2 - 5 MB | NPU
| Yolo-X | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 14.985 ms | 7 - 7 MB | NPU
| Yolo-X | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 8.841 ms | 2 - 207 MB | NPU
| Yolo-X | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 8.841 ms | 2 - 207 MB | NPU
| Yolo-X | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 11.068 ms | 0 - 208 MB | NPU
| Yolo-X | ONNX | w8a8 | Snapdragon® X2 Elite | 8.142 ms | 1 - 1 MB | NPU
| Yolo-X | ONNX | w8a8 | Snapdragon® X Elite | 8.936 ms | 8 - 8 MB | NPU
| Yolo-X | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 6.123 ms | 0 - 91 MB | NPU
| Yolo-X | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 8.9 ms | 1 - 92 MB | NPU
| Yolo-X | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 13.377 ms | 1 - 4 MB | NPU
| Yolo-X | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 8.262 ms | 0 - 47 MB | NPU
| Yolo-X | ONNX | w8a8 | Qualcomm® QCS8450 | 8.9 ms | 1 - 92 MB | NPU
| Yolo-X | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 10.254 ms | 1 - 4 MB | NPU
| Yolo-X | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 8.936 ms | 8 - 8 MB | NPU
| Yolo-X | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 5.784 ms | 1 - 60 MB | NPU
| Yolo-X | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 5.784 ms | 1 - 60 MB | NPU
| Yolo-X | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 8.091 ms | 0 - 62 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Snapdragon® X2 Elite | 28.4 ms | 16 - 16 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Snapdragon® X Elite | 35.987 ms | 19 - 19 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Gen 3 Mobile | 26.736 ms | 11 - 856 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 34.162 ms | 9 - 14 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ IQ-9075 | 36.44 ms | 11 - 14 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ IQ-X7181 | 35.987 ms | 19 - 19 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ Q-8750 | 23.45 ms | 12 - 806 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Elite Mobile | 23.45 ms | 12 - 806 MB | NPU
| Yolo-X | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Elite Gen 5 Mobile | 25.863 ms | 11 - 840 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 3.167 ms | 2 - 2 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® X Elite | 6.111 ms | 2 - 2 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 3.643 ms | 2 - 107 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 7.127 ms | 2 - 107 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 22.746 ms | 1 - 5 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8275 | 11.67 ms | 1 - 72 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.537 ms | 2 - 130 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® SA8775P | 6.135 ms | 0 - 74 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® SA8650P | 6.135 ms | 0 - 74 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® SA8255P | 6.135 ms | 0 - 74 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 7.127 ms | 2 - 107 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 5.866 ms | 1 - 5 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 6.111 ms | 2 - 2 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 46.205 ms | 3 - 196 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 6.893 ms | 2 - 190 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 2.606 ms | 2 - 81 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® SA7255P | 11.67 ms | 1 - 72 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Qualcomm® SA8295P | 7.584 ms | 0 - 69 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 2.606 ms | 2 - 81 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 1.961 ms | 2 - 83 MB | NPU
| Yolo-X | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 6.893 ms | 2 - 190 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 1.219 ms | 1 - 1 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® X Elite | 2.722 ms | 1 - 1 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.643 ms | 1 - 73 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 3.025 ms | 1 - 73 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 8.76 ms | 3 - 6 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 5.835 ms | 1 - 49 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.432 ms | 1 - 3 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® SA8775P | 2.834 ms | 1 - 53 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® SA8650P | 2.834 ms | 1 - 53 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® SA8255P | 2.834 ms | 1 - 53 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 3.025 ms | 1 - 73 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 2.538 ms | 1 - 4 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 2.722 ms | 1 - 1 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 21.637 ms | 1 - 162 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 2.791 ms | 1 - 49 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 1.175 ms | 1 - 52 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® SA7255P | 5.835 ms | 1 - 49 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Qualcomm® SA8295P | 3.787 ms | 1 - 48 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 1.175 ms | 1 - 52 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.911 ms | 1 - 53 MB | NPU
| Yolo-X | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 2.791 ms | 1 - 49 MB | NPU
| Yolo-X | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.654 ms | 0 - 77 MB | NPU
| Yolo-X | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 2.942 ms | 0 - 77 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 8.66 ms | 0 - 12 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 5.885 ms | 0 - 45 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.437 ms | 0 - 3 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® SA8775P | 2.883 ms | 0 - 51 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® SA8650P | 2.883 ms | 0 - 51 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® SA8255P | 2.883 ms | 0 - 51 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® QCS8450 | 2.942 ms | 0 - 77 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 2.638 ms | 0 - 12 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 21.686 ms | 0 - 159 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 2.776 ms | 0 - 45 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 1.205 ms | 0 - 53 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® SA7255P | 5.885 ms | 0 - 45 MB | NPU
| Yolo-X | TFLITE | w8a8 | Qualcomm® SA8295P | 3.834 ms | 0 - 45 MB | NPU
| Yolo-X | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 1.205 ms | 0 - 53 MB | NPU
| Yolo-X | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.92 ms | 0 - 50 MB | NPU
| Yolo-X | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 2.776 ms | 0 - 45 MB | NPU

## License
* The license for the original implementation of Yolo-X can be found
  [here](https://github.com/Megvii-BaseDetection/YOLOX/blob/main/LICENSE).

## References
* [YOLOX: Exceeding YOLO Series in 2021](https://github.com/Megvii-BaseDetection/YOLOX/blob/main/README.md)
* [Source Model Implementation](https://github.com/Megvii-BaseDetection/YOLOX/)

## 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).