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. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up 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 1.25.0 Download
ONNX w8a8 Universal QAIRT 2.45, ONNX Runtime 1.25.0 Download
QNN_DLC float Universal QAIRT 2.45 Download
QNN_DLC w8a8 Universal QAIRT 2.45 Download
TFLITE float Universal QAIRT 2.45 Download
TFLITE w8a8 Universal QAIRT 2.45 Download

For more device-specific assets and performance metrics, visit ResNet-3D on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models 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 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® X2 Elite 35.857 ms 12 - 12 MB NPU
ResNet-3D ONNX float Snapdragon® X Elite 67.783 ms 58 - 58 MB NPU
ResNet-3D ONNX float Snapdragon® 8 Gen 3 Mobile 50.647 ms 12 - 596 MB NPU
ResNet-3D ONNX float Snapdragon® 8 Gen 1 Mobile 133.711 ms 1 - 488 MB NPU
ResNet-3D ONNX float Qualcomm® Dragonwing™ QCS8550 (Proxy) 67.483 ms 0 - 70 MB NPU
ResNet-3D ONNX float Qualcomm® QCS8450 133.711 ms 1 - 488 MB NPU
ResNet-3D ONNX float Qualcomm® Dragonwing™ IQ-9075 123.178 ms 11 - 26 MB NPU
ResNet-3D ONNX float Snapdragon® 8 Elite Gen 5 Mobile 30.098 ms 1 - 407 MB NPU
ResNet-3D ONNX float Snapdragon® 8 Elite Mobile 41.705 ms 1 - 390 MB NPU
ResNet-3D ONNX float Qualcomm® Dragonwing™ Q-8750 41.705 ms 1 - 390 MB NPU
ResNet-3D ONNX float Qualcomm® Dragonwing™ IQ-X7181 67.783 ms 58 - 58 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® X2 Elite 11.557 ms 3 - 3 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® X Elite 22.419 ms 32 - 32 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® 8 Gen 3 Mobile 16.84 ms 0 - 413 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® 8 Gen 1 Mobile 36.196 ms 0 - 417 MB NPU
ResNet-3D ONNX w8a8 Qualcomm® Dragonwing™ QCS8550 (Proxy) 22.193 ms 0 - 40 MB NPU
ResNet-3D ONNX w8a8 Qualcomm® QCS8450 36.196 ms 0 - 417 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® 8 Elite Mobile 15.039 ms 0 - 293 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® 7 Gen 4 Mobile 42.16 ms 3 - 295 MB NPU
ResNet-3D ONNX w8a8 Snapdragon® 8 Elite Gen 5 Mobile 9.996 ms 2 - 300 MB NPU
ResNet-3D ONNX w8a8 Qualcomm® Dragonwing™ IQ-9075 21.313 ms 0 - 6 MB NPU
ResNet-3D ONNX w8a8 Qualcomm® Dragonwing™ Q-7790 42.16 ms 3 - 295 MB NPU
ResNet-3D ONNX w8a8 Qualcomm® Dragonwing™ Q-8750 15.039 ms 0 - 293 MB NPU
ResNet-3D ONNX w8a8 Qualcomm® Dragonwing™ IQ-X7181 22.419 ms 32 - 32 MB NPU
ResNet-3D QNN_DLC float Snapdragon® X2 Elite 36.361 ms 12 - 12 MB NPU
ResNet-3D QNN_DLC float Snapdragon® X Elite 70.616 ms 12 - 12 MB NPU
ResNet-3D QNN_DLC float Snapdragon® 8 Gen 3 Mobile 50.936 ms 12 - 600 MB NPU
ResNet-3D QNN_DLC float Snapdragon® 8 Gen 1 Mobile 134.755 ms 7 - 501 MB NPU
ResNet-3D QNN_DLC float Qualcomm® QCS8275 467.719 ms 0 - 396 MB NPU
ResNet-3D QNN_DLC float Qualcomm® Dragonwing™ QCS8550 (Proxy) 69.081 ms 12 - 48 MB NPU
ResNet-3D QNN_DLC float Qualcomm® SA8775P 123.936 ms 1 - 395 MB NPU
ResNet-3D QNN_DLC float Qualcomm® SA8650P 123.936 ms 1 - 395 MB NPU
ResNet-3D QNN_DLC float Qualcomm® SA8255P 123.936 ms 1 - 395 MB NPU
ResNet-3D QNN_DLC float Qualcomm® QCS8450 134.755 ms 7 - 501 MB NPU
ResNet-3D QNN_DLC float Qualcomm® Dragonwing™ IQ-9075 123.805 ms 13 - 27 MB NPU
ResNet-3D QNN_DLC float Snapdragon® 8 Elite Gen 5 Mobile 30.011 ms 12 - 406 MB NPU
ResNet-3D QNN_DLC float Qualcomm® SA7255P 467.719 ms 0 - 396 MB NPU
ResNet-3D QNN_DLC float Snapdragon® 8 Elite Mobile 41.732 ms 12 - 401 MB NPU
ResNet-3D QNN_DLC float Qualcomm® SA8295P 133.038 ms 0 - 317 MB NPU
ResNet-3D QNN_DLC float Qualcomm® Dragonwing™ Q-8750 41.732 ms 12 - 401 MB NPU
ResNet-3D QNN_DLC float Qualcomm® Dragonwing™ IQ-X7181 70.616 ms 12 - 12 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® X2 Elite 12.375 ms 3 - 3 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® X Elite 23.398 ms 3 - 3 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® 8 Gen 3 Mobile 17.098 ms 3 - 411 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® 8 Gen 1 Mobile 38.046 ms 3 - 413 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ QCS6490 137.01 ms 3 - 7 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® QCS8275 75.863 ms 3 - 284 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ QCS8550 (Proxy) 22.729 ms 3 - 5 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® SA8775P 21.813 ms 3 - 284 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® SA8650P 21.813 ms 3 - 284 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® SA8255P 21.813 ms 3 - 284 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® QCS8450 38.046 ms 3 - 413 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® 8 Elite Mobile 15.452 ms 3 - 292 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® 7 Gen 4 Mobile 43.297 ms 3 - 300 MB NPU
ResNet-3D QNN_DLC w8a8 Snapdragon® 8 Elite Gen 5 Mobile 10.672 ms 3 - 299 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ IQ-9075 21.073 ms 5 - 9 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ Q-6690 472.175 ms 6 - 363 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® SA7255P 75.863 ms 3 - 284 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® SA8295P 39.827 ms 0 - 280 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ Q-7790 43.297 ms 3 - 300 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ Q-8750 15.452 ms 3 - 292 MB NPU
ResNet-3D QNN_DLC w8a8 Qualcomm® Dragonwing™ IQ-X7181 23.398 ms 3 - 3 MB NPU
ResNet-3D TFLITE float Snapdragon® 8 Gen 3 Mobile 1069.241 ms 0 - 636 MB NPU
ResNet-3D TFLITE float Snapdragon® 8 Gen 1 Mobile 1625.325 ms 1 - 606 MB NPU
ResNet-3D TFLITE float Qualcomm® QCS8275 2774.41 ms 1 - 457 MB NPU
ResNet-3D TFLITE float Qualcomm® Dragonwing™ QCS8550 (Proxy) 1474.312 ms 0 - 4 MB NPU
ResNet-3D TFLITE float Qualcomm® SA8775P 1478.637 ms 1 - 476 MB NPU
ResNet-3D TFLITE float Qualcomm® SA8650P 1478.637 ms 1 - 476 MB NPU
ResNet-3D TFLITE float Qualcomm® SA8255P 1478.637 ms 1 - 476 MB NPU
ResNet-3D TFLITE float Qualcomm® QCS8450 1625.325 ms 1 - 606 MB NPU
ResNet-3D TFLITE float Qualcomm® Dragonwing™ IQ-9075 1418.372 ms 0 - 83 MB NPU
ResNet-3D TFLITE float Snapdragon® 8 Elite Gen 5 Mobile 885.903 ms 1 - 473 MB NPU
ResNet-3D TFLITE float Qualcomm® SA7255P 2774.41 ms 1 - 457 MB NPU
ResNet-3D TFLITE float Snapdragon® 8 Elite Mobile 988.491 ms 0 - 466 MB NPU
ResNet-3D TFLITE float Qualcomm® SA8295P 1734.663 ms 0 - 454 MB NPU
ResNet-3D TFLITE float Qualcomm® Dragonwing™ Q-8750 988.491 ms 0 - 466 MB NPU
ResNet-3D TFLITE w8a8 Snapdragon® 8 Gen 3 Mobile 934.855 ms 0 - 1008 MB NPU
ResNet-3D TFLITE w8a8 Snapdragon® 8 Gen 1 Mobile 1413.153 ms 0 - 867 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® Dragonwing™ QCS6490 9658.683 ms 3462 - 5278 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® QCS8275 2556.009 ms 1 - 820 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® Dragonwing™ QCS8550 (Proxy) 1256.34 ms 0 - 4 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® SA8775P 1259.878 ms 1 - 819 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® SA8650P 1259.878 ms 1 - 819 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® SA8255P 1259.878 ms 1 - 819 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® QCS8450 1413.153 ms 0 - 867 MB NPU
ResNet-3D TFLITE w8a8 Snapdragon® 8 Elite Mobile 681.809 ms 1 - 724 MB NPU
ResNet-3D TFLITE w8a8 Snapdragon® 7 Gen 4 Mobile 341128.645 ms 920 - 1201 MB NPU
ResNet-3D TFLITE w8a8 Snapdragon® 8 Elite Gen 5 Mobile 820.336 ms 0 - 926 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® Dragonwing™ IQ-9075 1272.405 ms 0 - 77 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® Dragonwing™ Q-6690 384880.936 ms 953 - 1169 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® SA7255P 2556.009 ms 1 - 820 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® SA8295P 1342.174 ms 1 - 685 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® Dragonwing™ Q-7790 341128.645 ms 920 - 1201 MB NPU
ResNet-3D TFLITE w8a8 Qualcomm® Dragonwing™ Q-8750 681.809 ms 1 - 724 MB NPU

License

  • The license for the original implementation of ResNet-3D can be found here.

References

Community

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for qualcomm/ResNet-3D