README.md

---
library_name: pytorch
license: other
tags:
- backbone
- android
pipeline_tag: video-classification

---

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

# ResNet-Mixed-Convolution: Optimized for Qualcomm Devices

ResNet Mixed Convolutions is a network with a mixture of 2D and 3D convolutions used for video understanding.

This is based on the implementation of ResNet-Mixed-Convolution found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/video/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/main/qai_hub_models/models/resnet_mixed) 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 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_mixed/releases/v0.48.0/resnet_mixed-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_mixed/releases/v0.48.0/resnet_mixed-onnx-w8a16.zip)
| QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_mixed/releases/v0.48.0/resnet_mixed-qnn_dlc-float.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_mixed/releases/v0.48.0/resnet_mixed-qnn_dlc-w8a16.zip)
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_mixed/releases/v0.48.0/resnet_mixed-tflite-float.zip)

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


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/resnet_mixed) 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 [ResNet-Mixed-Convolution on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/resnet_mixed) for usage instructions.

## Model Details

**Model Type:** Model_use_case.video_classification

**Model Stats:**
- Model checkpoint: Kinetics-400
- Input resolution: 112x112
- Number of parameters: 11.7M
- Model size (float): 44.6 MB
- Model size (w8a16): 11.5 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ResNet-Mixed-Convolution | ONNX | float | Snapdragon® X2 Elite | 7.099 ms | 22 - 22 MB | NPU
| ResNet-Mixed-Convolution | ONNX | float | Snapdragon® X Elite | 13.823 ms | 22 - 22 MB | NPU
| ResNet-Mixed-Convolution | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 9.611 ms | 2 - 276 MB | NPU
| ResNet-Mixed-Convolution | ONNX | float | Qualcomm® QCS8550 (Proxy) | 13.288 ms | 0 - 29 MB | NPU
| ResNet-Mixed-Convolution | ONNX | float | Qualcomm® QCS9075 | 26.624 ms | 2 - 5 MB | NPU
| ResNet-Mixed-Convolution | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 7.811 ms | 0 - 215 MB | NPU
| ResNet-Mixed-Convolution | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.559 ms | 0 - 219 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Snapdragon® X2 Elite | 4.647 ms | 12 - 12 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Snapdragon® X Elite | 9.09 ms | 11 - 11 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 6.392 ms | 1 - 252 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Qualcomm® QCS6490 | 1727.214 ms | 51 - 62 MB | CPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 8.515 ms | 1 - 5 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Qualcomm® QCS9075 | 9.054 ms | 1 - 4 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Qualcomm® QCM6690 | 890.657 ms | 107 - 114 MB | CPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 5.285 ms | 1 - 193 MB | NPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 874.466 ms | 112 - 120 MB | CPU
| ResNet-Mixed-Convolution | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.516 ms | 0 - 190 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Snapdragon® X2 Elite | 7.652 ms | 2 - 2 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Snapdragon® X Elite | 13.982 ms | 2 - 2 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 9.56 ms | 0 - 284 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 97.146 ms | 1 - 223 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 13.22 ms | 2 - 34 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® SA8775P | 25.075 ms | 1 - 222 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® QCS9075 | 27.603 ms | 2 - 6 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 27.484 ms | 2 - 244 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® SA7255P | 97.146 ms | 1 - 223 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Qualcomm® SA8295P | 26.782 ms | 0 - 194 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 7.629 ms | 0 - 225 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.732 ms | 2 - 233 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 5.359 ms | 1 - 1 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Snapdragon® X Elite | 9.823 ms | 1 - 1 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 6.624 ms | 1 - 254 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 37.274 ms | 0 - 3 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 29.75 ms | 1 - 190 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 9.172 ms | 1 - 3 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® SA8775P | 9.393 ms | 1 - 191 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 10.391 ms | 1 - 4 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 182.729 ms | 1 - 208 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 14.0 ms | 1 - 255 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® SA7255P | 29.75 ms | 1 - 190 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Qualcomm® SA8295P | 16.32 ms | 1 - 192 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 5.267 ms | 1 - 187 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 15.899 ms | 1 - 197 MB | NPU
| ResNet-Mixed-Convolution | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 4.005 ms | 7 - 195 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 217.382 ms | 0 - 312 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 584.652 ms | 0 - 249 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 304.587 ms | 0 - 2 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Qualcomm® SA8775P | 294.794 ms | 0 - 249 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 339.962 ms | 0 - 278 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Qualcomm® SA7255P | 584.652 ms | 0 - 249 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Qualcomm® SA8295P | 363.01 ms | 0 - 232 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 201.863 ms | 0 - 246 MB | NPU
| ResNet-Mixed-Convolution | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 198.096 ms | 0 - 251 MB | NPU

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

## References
* [A Closer Look at Spatiotemporal Convolutions for Action Recognition](https://arxiv.org/abs/1711.11248)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/video/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).