ResNet-2Plus1D: Optimized for Qualcomm Devices

ResNet (2+1)D Convolutions is a network which explicitly factorizes 3D convolution into two separate and successive operations, a 2D spatial convolution and a 1D temporal convolution. It used for video understanding applications.

This is based on the implementation of ResNet-2Plus1D 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 1.27.1	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	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-2Plus1D 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-2Plus1D on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.video_classification

Model Stats:

• Input resolution: 112x112
• Model checkpoint: Kinetics-400
• Model size (float): 120 MB
• Model size (w8a8): 30.8 MB
• Number of parameters: 31.5M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNet-2Plus1D	ONNX	float	Snapdragon® X2 Elite	34.001 ms	12 - 12 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® X Elite	66.983 ms	57 - 57 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	52.811 ms	4 - 767 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Gen 1 Mobile	146.499 ms	13 - 601 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	125.259 ms	11 - 26 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	69.752 ms	12 - 15 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS8450	146.499 ms	13 - 601 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	111.302 ms	11 - 26 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	66.983 ms	57 - 57 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® Dragonwing™ Q-8750	42.361 ms	2 - 550 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite Mobile	42.361 ms	2 - 550 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	32.055 ms	1 - 569 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X2 Elite	12.0 ms	3 - 3 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X Elite	24.988 ms	32 - 32 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	19.644 ms	0 - 467 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	44.569 ms	0 - 459 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-8275	25.442 ms	0 - 7 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	25.166 ms	3 - 6 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS8450	44.569 ms	0 - 459 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-9075	23.9 ms	0 - 6 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-X7181	24.988 ms	32 - 32 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® Dragonwing™ Q-7790	42.456 ms	3 - 388 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® Dragonwing™ Q-8750	15.748 ms	0 - 372 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite Mobile	15.748 ms	0 - 372 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	12.053 ms	0 - 386 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	42.456 ms	3 - 388 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X2 Elite	34.44 ms	12 - 12 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X Elite	66.569 ms	12 - 12 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	52.056 ms	12 - 721 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	143.861 ms	3 - 593 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	126.424 ms	12 - 25 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	424.55 ms	1 - 492 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	70.321 ms	12 - 620 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8775P	111.025 ms	0 - 489 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8650P	111.025 ms	0 - 489 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8255P	111.025 ms	0 - 489 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8450	143.861 ms	3 - 593 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	111.631 ms	12 - 25 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	66.569 ms	12 - 12 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	41.671 ms	12 - 498 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA7255P	424.55 ms	1 - 492 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8295P	118.419 ms	0 - 398 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite Mobile	41.671 ms	12 - 498 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	29.877 ms	12 - 509 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X2 Elite	13.524 ms	3 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X Elite	26.976 ms	3 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	21.212 ms	3 - 466 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	46.779 ms	3 - 460 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS6490	131.195 ms	3 - 7 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-8275	27.258 ms	3 - 8 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-8275	75.078 ms	3 - 373 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	27.949 ms	3 - 6 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8775P	24.5 ms	3 - 372 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8650P	24.5 ms	3 - 372 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8255P	24.5 ms	3 - 372 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8450	46.779 ms	3 - 460 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	24.717 ms	1 - 6 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	26.976 ms	3 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-6690	439.335 ms	3 - 425 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-7790	46.448 ms	3 - 389 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	16.813 ms	3 - 373 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA7255P	75.078 ms	3 - 373 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8295P	41.852 ms	0 - 367 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	16.813 ms	3 - 373 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	13.252 ms	3 - 390 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	46.448 ms	3 - 389 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1491.858 ms	0 - 801 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	2382.786 ms	3 - 739 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	1926.153 ms	0 - 82 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	3623.388 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1959.927 ms	0 - 3 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8775P	1953.551 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8650P	1953.551 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8255P	1953.551 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8450	2382.786 ms	3 - 739 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	2002.328 ms	0 - 82 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1408.387 ms	0 - 576 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA7255P	3623.388 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8295P	2470.694 ms	1 - 540 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite Mobile	1408.387 ms	0 - 576 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1276.429 ms	0 - 573 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	3124.592 ms	1 - 1618 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	4715.281 ms	3 - 1341 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS6490	8665.541 ms	1395 - 2187 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-8275	4072.248 ms	0 - 80 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-8275	8129.711 ms	3 - 1383 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4142.206 ms	1 - 5 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8775P	4291.826 ms	2 - 1381 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8650P	4291.826 ms	2 - 1381 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8255P	4291.826 ms	2 - 1381 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8450	4715.281 ms	3 - 1341 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-9075	4299.918 ms	0 - 79 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-6690	8055.421 ms	1334 - 1577 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-7790	6152.013 ms	1324 - 1706 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-8750	2855.741 ms	1 - 1675 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA7255P	8129.711 ms	3 - 1383 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8295P	4658.245 ms	3 - 1092 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	2855.741 ms	1 - 1675 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3451.611 ms	0 - 1453 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	6152.013 ms	1324 - 1706 MB	NPU

License

• The license for the original implementation of ResNet-2Plus1D can be found here.

References

• A Closer Look at Spatiotemporal Convolutions for Action Recognition
• Source Model Implementation

Community

• Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
• For questions or feedback please reach out to us.