---
library_name: pytorch
license: bsd-3-clause
tags:
- backbone
- bu_auto
- android
pipeline_tag: image-classification

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/web-assets/model_demo.png)

# ResNet50: Optimized for Qualcomm Devices

ResNet50 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.

This is based on the implementation of ResNet50 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py).
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/v0.63.0/src/qai_hub_models/models/resnet50) 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 2.50, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.63.0/resnet50-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.50, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.63.0/resnet50-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.63.0/resnet50-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.63.0/resnet50-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.63.0/resnet50-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.63.0/resnet50-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[ResNet50 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/resnet50)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.63.0/src/qai_hub_models/models/resnet50) 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 [ResNet50 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.63.0/src/qai_hub_models/models/resnet50) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Input resolution: 224x224
- Model checkpoint: Imagenet
- Model size (float): 97.4 MB
- Model size (w8a8): 25.1 MB
- Number of parameters: 25.5M

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ResNet50 | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.066 ms | 0 - 51 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.203 ms | 0 - 50 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® X2 Elite | 0.968 ms | 2 - 2 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® X Elite | 1.917 ms | 50 - 50 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.433 ms | 0 - 85 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 2.882 ms | 0 - 68 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 3.314 ms | 0 - 5 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.988 ms | 0 - 58 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® QCS8450 | 2.882 ms | 0 - 68 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 3.008 ms | 1 - 4 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.917 ms | 50 - 50 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.203 ms | 0 - 50 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.554 ms | 0 - 53 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.58 ms | 0 - 53 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® X2 Elite | 0.408 ms | 1 - 1 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® X Elite | 0.842 ms | 25 - 25 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.683 ms | 0 - 91 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.059 ms | 0 - 89 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 2.666 ms | 0 - 3 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.98 ms | 0 - 4 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.904 ms | 0 - 160 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® QCS8450 | 1.059 ms | 0 - 89 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.009 ms | 0 - 3 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.842 ms | 25 - 25 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 6.237 ms | 0 - 169 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.148 ms | 0 - 58 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.58 ms | 0 - 53 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.148 ms | 0 - 58 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.098 ms | 1 - 52 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.284 ms | 1 - 48 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® X2 Elite | 1.198 ms | 1 - 1 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® X Elite | 2.398 ms | 1 - 1 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.615 ms | 0 - 78 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 3.469 ms | 1 - 67 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 3.359 ms | 1 - 4 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.244 ms | 1 - 2 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8775P | 3.355 ms | 1 - 50 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8650P | 3.355 ms | 1 - 50 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8255P | 3.355 ms | 1 - 50 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® QCS8450 | 3.469 ms | 1 - 67 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 3.163 ms | 3 - 5 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.398 ms | 1 - 1 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.284 ms | 1 - 48 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA7255P | 10.671 ms | 1 - 48 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8295P | 3.557 ms | 0 - 32 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.487 ms | 0 - 49 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.529 ms | 0 - 47 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.461 ms | 0 - 0 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.945 ms | 0 - 0 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.658 ms | 0 - 77 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.052 ms | 0 - 83 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.007 ms | 0 - 2 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.931 ms | 0 - 3 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.901 ms | 0 - 1 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.064 ms | 0 - 52 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.064 ms | 0 - 52 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.064 ms | 0 - 52 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.052 ms | 0 - 83 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.961 ms | 2 - 4 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.945 ms | 0 - 0 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 6.444 ms | 2 - 163 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.225 ms | 0 - 55 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.529 ms | 0 - 47 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA7255P | 1.968 ms | 0 - 50 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8295P | 1.298 ms | 0 - 47 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.225 ms | 0 - 55 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.117 ms | 0 - 73 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.26 ms | 0 - 71 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.602 ms | 0 - 117 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 3.48 ms | 0 - 102 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 3.354 ms | 0 - 53 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.202 ms | 0 - 2 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® SA8775P | 3.297 ms | 0 - 73 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® SA8650P | 3.297 ms | 0 - 73 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® SA8255P | 3.297 ms | 0 - 73 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® QCS8450 | 3.48 ms | 0 - 102 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 3.139 ms | 0 - 52 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.26 ms | 0 - 71 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® SA8295P | 3.488 ms | 0 - 60 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.449 ms | 0 - 51 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.478 ms | 0 - 43 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.571 ms | 0 - 79 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.92 ms | 0 - 80 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 2.565 ms | 0 - 27 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.837 ms | 0 - 28 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.755 ms | 0 - 4 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® SA8775P | 0.955 ms | 0 - 50 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® SA8650P | 0.955 ms | 0 - 50 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® SA8255P | 0.955 ms | 0 - 50 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® QCS8450 | 0.92 ms | 0 - 80 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.841 ms | 0 - 27 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 5.924 ms | 0 - 158 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.083 ms | 0 - 52 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.478 ms | 0 - 43 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® SA8295P | 1.152 ms | 0 - 44 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.083 ms | 0 - 52 MB | NPU

## License
* The license for the original implementation of ResNet50 can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).

## References
* [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py)

## 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).