---
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/resnext50/web-assets/model_demo.png)

# ResNeXt50: Optimized for Qualcomm Devices

ResNeXt50 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 ResNeXt50 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/v0.65.0/src/qai_hub_models/models/resnext50) 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 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnext50/releases/v0.65.0/resnext50-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.50, ONNX Runtime 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnext50/releases/v0.65.0/resnext50-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/resnext50/releases/v0.65.0/resnext50-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/resnext50/releases/v0.65.0/resnext50-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/resnext50/releases/v0.65.0/resnext50-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnext50/releases/v0.65.0/resnext50-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[ResNeXt50 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/resnext50)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.65.0/src/qai_hub_models/models/resnext50) 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 [ResNeXt50 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.65.0/src/qai_hub_models/models/resnext50) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Input resolution: 224x224
- Model checkpoint: Imagenet
- Model size (float): 95.4 MB
- Model size (w8a8): 24.8 MB
- Number of parameters: 25.0M

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ResNeXt50 | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.176 ms | 1 - 76 MB | NPU
| ResNeXt50 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.376 ms | 0 - 76 MB | NPU
| ResNeXt50 | ONNX | float | Snapdragon® X2 Elite | 1.108 ms | 2 - 2 MB | NPU
| ResNeXt50 | ONNX | float | Snapdragon® X Elite | 2.204 ms | 50 - 50 MB | NPU
| ResNeXt50 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.623 ms | 0 - 140 MB | NPU
| ResNeXt50 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 3.221 ms | 0 - 119 MB | NPU
| ResNeXt50 | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 3.791 ms | 0 - 5 MB | NPU
| ResNeXt50 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.286 ms | 0 - 58 MB | NPU
| ResNeXt50 | ONNX | float | Qualcomm® QCS8450 | 3.221 ms | 0 - 119 MB | NPU
| ResNeXt50 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 3.439 ms | 0 - 4 MB | NPU
| ResNeXt50 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.204 ms | 50 - 50 MB | NPU
| ResNeXt50 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.376 ms | 0 - 76 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.619 ms | 0 - 84 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.69 ms | 0 - 82 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® X2 Elite | 0.512 ms | 1 - 1 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® X Elite | 1.055 ms | 25 - 25 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.805 ms | 0 - 109 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.311 ms | 0 - 111 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.499 ms | 0 - 3 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 1.19 ms | 0 - 4 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.087 ms | 0 - 30 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® QCS8450 | 1.311 ms | 0 - 111 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.316 ms | 0 - 3 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.055 ms | 25 - 25 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 7.409 ms | 0 - 195 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.284 ms | 0 - 85 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.69 ms | 0 - 82 MB | NPU
| ResNeXt50 | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.284 ms | 0 - 85 MB | NPU
| ResNeXt50 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.216 ms | 1 - 76 MB | NPU
| ResNeXt50 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.465 ms | 0 - 75 MB | NPU
| ResNeXt50 | QNN_DLC | float | Snapdragon® X2 Elite | 1.428 ms | 1 - 1 MB | NPU
| ResNeXt50 | QNN_DLC | float | Snapdragon® X Elite | 2.671 ms | 1 - 1 MB | NPU
| ResNeXt50 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.806 ms | 0 - 140 MB | NPU
| ResNeXt50 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 3.842 ms | 0 - 118 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 3.766 ms | 1 - 4 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.483 ms | 1 - 97 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® SA8650P | 3.733 ms | 1 - 75 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® SA8255P | 3.733 ms | 1 - 75 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® QCS8450 | 3.842 ms | 0 - 118 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 3.568 ms | 1 - 3 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.671 ms | 1 - 1 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.465 ms | 0 - 75 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® SA7255P | 11.864 ms | 1 - 74 MB | NPU
| ResNeXt50 | QNN_DLC | float | Qualcomm® SA8295P | 3.952 ms | 0 - 55 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.54 ms | 0 - 78 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.622 ms | 0 - 75 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.614 ms | 0 - 0 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.187 ms | 0 - 0 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.787 ms | 0 - 106 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.315 ms | 0 - 108 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.44 ms | 0 - 2 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 1.134 ms | 0 - 3 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.041 ms | 0 - 8 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.4 ms | 0 - 79 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.4 ms | 0 - 79 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.4 ms | 0 - 79 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.315 ms | 0 - 108 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.272 ms | 0 - 2 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.187 ms | 0 - 0 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 7.751 ms | 0 - 192 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.357 ms | 0 - 82 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.622 ms | 0 - 75 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® SA7255P | 2.361 ms | 0 - 77 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Qualcomm® SA8295P | 1.605 ms | 0 - 76 MB | NPU
| ResNeXt50 | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.357 ms | 0 - 82 MB | NPU
| ResNeXt50 | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.203 ms | 0 - 120 MB | NPU
| ResNeXt50 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.439 ms | 0 - 116 MB | NPU
| ResNeXt50 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.752 ms | 0 - 178 MB | NPU
| ResNeXt50 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 3.798 ms | 0 - 157 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 3.797 ms | 0 - 54 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.48 ms | 0 - 2 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® SA8650P | 3.791 ms | 0 - 116 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® SA8255P | 3.791 ms | 0 - 116 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® QCS8450 | 3.798 ms | 0 - 157 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 3.573 ms | 0 - 53 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.439 ms | 0 - 116 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® SA7255P | 11.923 ms | 0 - 113 MB | NPU
| ResNeXt50 | TFLITE | float | Qualcomm® SA8295P | 4.007 ms | 0 - 90 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.495 ms | 0 - 69 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.56 ms | 0 - 67 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.66 ms | 0 - 104 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.123 ms | 0 - 103 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.132 ms | 0 - 27 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.99 ms | 0 - 28 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.909 ms | 0 - 5 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® SA8775P | 1.212 ms | 0 - 74 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® SA8650P | 1.212 ms | 0 - 74 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® SA8255P | 1.212 ms | 0 - 74 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® QCS8450 | 1.123 ms | 0 - 103 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.07 ms | 0 - 28 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 7.284 ms | 0 - 183 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.224 ms | 0 - 73 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.56 ms | 0 - 67 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Qualcomm® SA8295P | 1.41 ms | 0 - 68 MB | NPU
| ResNeXt50 | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.224 ms | 0 - 73 MB | NPU

## License
* The license for the original implementation of ResNeXt50 can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).

## References
* [Aggregated Residual Transformations for Deep Neural Networks](https://arxiv.org/abs/1611.05431)
* [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).