---
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_3d/web-assets/model_demo.png)

# ResNet-3D: Optimized for Qualcomm Devices

ResNet 3D is a network with 3D convolutions used for video understanding.

This is based on the implementation of ResNet-3D 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/src/qai_hub_models/models/resnet_3d) 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 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_3d/releases/v0.55.0/resnet_3d-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_3d/releases/v0.55.0/resnet_3d-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_3d/releases/v0.55.0/resnet_3d-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_3d/releases/v0.55.0/resnet_3d-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_3d/releases/v0.55.0/resnet_3d-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet_3d/releases/v0.55.0/resnet_3d-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[ResNet-3D on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/resnet_3d)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/resnet_3d) 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-3D on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/resnet_3d) for usage instructions.

## Model Details

**Model Type:** Model_use_case.video_classification

**Model Stats:**
- Model checkpoint: Kinetics-400
- Input resolution: 112x112
- Number of parameters: 33.4M
- Model size (float): 127 MB
- Model size (w8a8): 32.1 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ResNet-3D | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 30.102 ms | 12 - 406 MB | NPU
| ResNet-3D | ONNX | float | Snapdragon® X Elite | 69.454 ms | 170 - 170 MB | NPU
| ResNet-3D | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 50.522 ms | 12 - 624 MB | NPU
| ResNet-3D | ONNX | float | Qualcomm® QCS8550 (Proxy) | 68.545 ms | 0 - 85 MB | NPU
| ResNet-3D | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 41.592 ms | 1 - 384 MB | NPU
| ResNet-3D | ONNX | float | Qualcomm® QCS9075 | 129.624 ms | 11 - 68 MB | NPU
| ResNet-3D | ONNX | float | Qualcomm® QCS8750 | 41.592 ms | 1 - 384 MB | NPU
| ResNet-3D | ONNX | float | Qualcomm® QCS7181 | 69.454 ms | 170 - 170 MB | NPU
| ResNet-3D | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 10.153 ms | 3 - 291 MB | NPU
| ResNet-3D | ONNX | w8a8 | Snapdragon® X2 Elite | 11.391 ms | 209 - 209 MB | NPU
| ResNet-3D | ONNX | w8a8 | Snapdragon® X Elite | 22.495 ms | 179 - 179 MB | NPU
| ResNet-3D | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 16.798 ms | 3 - 427 MB | NPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCS6490 | 1297.538 ms | 168 - 243 MB | CPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 22.293 ms | 3 - 71 MB | NPU
| ResNet-3D | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1482.241 ms | 64 - 71 MB | CPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCM6690 | 1581.396 ms | 54 - 61 MB | CPU
| ResNet-3D | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 15.003 ms | 0 - 288 MB | NPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCS9075 | 22.024 ms | 3 - 48 MB | NPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCS7790 | 1482.241 ms | 64 - 71 MB | CPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCS8750 | 15.003 ms | 0 - 288 MB | NPU
| ResNet-3D | ONNX | w8a8 | Qualcomm® QCS7181 | 22.495 ms | 179 - 179 MB | NPU
| ResNet-3D | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 30.02 ms | 12 - 401 MB | NPU
| ResNet-3D | QNN_DLC | float | Snapdragon® X2 Elite | 36.466 ms | 12 - 12 MB | NPU
| ResNet-3D | QNN_DLC | float | Snapdragon® X Elite | 70.666 ms | 12 - 12 MB | NPU
| ResNet-3D | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 51.057 ms | 12 - 599 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® QCS8275 | 467.94 ms | 1 - 396 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 68.8 ms | 12 - 14 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® SA8775P | 123.865 ms | 0 - 395 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® SA8650P | 123.865 ms | 0 - 395 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® SA8255P | 123.865 ms | 0 - 395 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 134.581 ms | 8 - 501 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® SA8295P | 133.01 ms | 0 - 317 MB | NPU
| ResNet-3D | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 41.947 ms | 10 - 399 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® SA7255P | 467.94 ms | 1 - 396 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® QCS9075 | 130.73 ms | 13 - 27 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® QCS8750 | 41.947 ms | 10 - 399 MB | NPU
| ResNet-3D | QNN_DLC | float | Qualcomm® QCS7181 | 70.666 ms | 12 - 12 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 10.631 ms | 3 - 295 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 12.276 ms | 3 - 3 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Snapdragon® X Elite | 23.422 ms | 3 - 3 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 17.08 ms | 3 - 409 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 102.251 ms | 2 - 7 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS8275 | 75.887 ms | 3 - 283 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 22.739 ms | 3 - 6 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® SA8775P | 21.847 ms | 3 - 284 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® SA8650P | 21.847 ms | 3 - 284 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® SA8255P | 21.847 ms | 3 - 284 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® SA7255P | 75.887 ms | 3 - 283 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 43.335 ms | 3 - 301 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® SA8295P | 39.779 ms | 0 - 280 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 453.961 ms | 3 - 358 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 15.619 ms | 3 - 290 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 22.439 ms | 2 - 7 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 36.485 ms | 3 - 410 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS7790 | 43.335 ms | 3 - 301 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS8750 | 15.619 ms | 3 - 290 MB | NPU
| ResNet-3D | QNN_DLC | w8a8 | Qualcomm® QCS7181 | 23.422 ms | 3 - 3 MB | NPU
| ResNet-3D | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 895.502 ms | 0 - 463 MB | NPU
| ResNet-3D | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1136.629 ms | 0 - 647 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 1476.595 ms | 0 - 95 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® SA8775P | 1479.959 ms | 1 - 476 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® SA8650P | 1479.959 ms | 1 - 476 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® SA8255P | 1479.959 ms | 1 - 476 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1645.363 ms | 0 - 602 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® SA8295P | 1736.112 ms | 0 - 455 MB | NPU
| ResNet-3D | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 965.516 ms | 0 - 470 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® QCS9075 | 1457.224 ms | 0 - 82 MB | NPU
| ResNet-3D | TFLITE | float | Qualcomm® QCS8750 | 965.516 ms | 0 - 470 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 818.797 ms | 0 - 916 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 950.813 ms | 0 - 1003 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS6490 | 10336.472 ms | 3462 - 5278 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS8275 | 2553.072 ms | 1 - 818 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 1254.704 ms | 0 - 4 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® SA8775P | 1257.932 ms | 0 - 819 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® SA8650P | 1257.932 ms | 0 - 819 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® SA8255P | 1257.932 ms | 0 - 819 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® SA7255P | 2553.072 ms | 1 - 818 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 343377.863 ms | 919 - 1200 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® SA8295P | 1343.007 ms | 1 - 685 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCM6690 | 378500.381 ms | 905 - 1116 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 646.203 ms | 1 - 723 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS9075 | 1115.378 ms | 0 - 74 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 1431.716 ms | 0 - 867 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS7790 | 343377.863 ms | 919 - 1200 MB | NPU
| ResNet-3D | TFLITE | w8a8 | Qualcomm® QCS8750 | 646.203 ms | 1 - 723 MB | NPU

## License
* The license for the original implementation of ResNet-3D 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).