BEVFusion: Optimized for Qualcomm Devices

BeVFusion is a machine learning model for generating a birds eye view represenation from the sensors(cameras) mounted on a vehicle.

This is based on the implementation of BEVFusion 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
PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	QAIRT 2.42, ONNX Runtime 1.25.0	Download
PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	QAIRT 2.42, ONNX Runtime 1.25.0	Download
PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	QAIRT 2.42, ONNX Runtime 1.25.0	Download
PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	QAIRT 2.42, ONNX Runtime 1.25.0	Download
PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	QAIRT 2.42, ONNX Runtime 1.25.0	Download
QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Qualcomm® SA8775P	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Qualcomm® SA7255P	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	QAIRT 2.45	Download
QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit BEVFusion 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 BEVFusion on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.driver_assistance

Model Stats:

Model checkpoint: camera-only-det.pth
Input resolution: 1 x 6 x 3 x 256 x 704
Number of parameters: 44M
Model size: 171 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	5.747 ms	16 - 26 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	7.034 ms	22 - 22 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.797 ms	16 - 22 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.642 ms	13 - 20 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	24.614 ms	14 - 17 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8750	7.642 ms	13 - 20 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	5.721 ms	5 - 14 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	7.425 ms	5 - 5 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	13.203 ms	5 - 5 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	9.813 ms	5 - 12 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275	92.822 ms	1 - 10 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	12.75 ms	5 - 6 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® SA8775P	23.782 ms	0 - 10 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® SA8650P	23.782 ms	0 - 10 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® SA8255P	23.782 ms	0 - 10 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	26.04 ms	0 - 6 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	7.736 ms	5 - 18 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	24.293 ms	7 - 18 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® SA7255P	92.822 ms	1 - 10 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8750	7.736 ms	5 - 18 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS7181	13.203 ms	5 - 5 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	345.766 ms	51 - 60 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	382.932 ms	66 - 66 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	530.944 ms	32 - 44 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	423.4 ms	20 - 32 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	706.111 ms	12 - 27 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8750	423.4 ms	20 - 32 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	354.115 ms	11 - 20 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	397.011 ms	12 - 12 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	702.28 ms	12 - 12 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	522.015 ms	13 - 19 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275	1071.609 ms	1 - 9 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	724.273 ms	13 - 14 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® SA8775P	719.921 ms	0 - 10 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® SA8650P	719.921 ms	0 - 10 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® SA8255P	719.921 ms	0 - 10 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	815.75 ms	1 - 6 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	440.382 ms	0 - 13 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	704.627 ms	12 - 43 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® SA7255P	1071.609 ms	1 - 9 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8750	440.382 ms	0 - 13 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS7181	702.28 ms	12 - 12 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2199.096 ms	49 - 58 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	2388.091 ms	1058 - 1058 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	2719.401 ms	442 - 449 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	2447.711 ms	556 - 564 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	3592.493 ms	602 - 604 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8750	2447.711 ms	556 - 564 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	2251.474 ms	17 - 26 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	2338.424 ms	17 - 17 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	3425.543 ms	17 - 17 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	2796.234 ms	17 - 27 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275	5463.389 ms	16 - 25 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	3481.367 ms	17 - 20 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® SA8775P	3415.752 ms	17 - 27 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® SA8650P	3415.752 ms	17 - 27 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® SA8255P	3415.752 ms	17 - 27 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	3683.038 ms	1 - 12 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	2438.52 ms	3 - 16 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	3405.08 ms	17 - 644 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® SA7255P	5463.389 ms	16 - 25 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8750	2438.52 ms	3 - 16 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS7181	3425.543 ms	17 - 17 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.78 ms	29 - 39 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	7.81 ms	19 - 19 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.121 ms	31 - 38 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.74 ms	14 - 21 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	15.16 ms	18 - 39 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8750	7.74 ms	14 - 21 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	6.785 ms	18 - 28 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	8.342 ms	18 - 18 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	12.137 ms	18 - 18 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	9.29 ms	19 - 26 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275	25.262 ms	18 - 26 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	12.582 ms	18 - 20 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® SA8775P	15.166 ms	18 - 27 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® SA8650P	15.166 ms	18 - 27 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® SA8255P	15.166 ms	18 - 27 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® SA8295P	14.937 ms	0 - 6 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	7.712 ms	18 - 26 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	14.815 ms	18 - 43 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® SA7255P	25.262 ms	18 - 26 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8750	7.712 ms	18 - 26 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS7181	12.137 ms	18 - 18 MB	NPU

License

The license for the original implementation of BEVFusion can be found here.

References

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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

Other

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for qualcomm/BEVFusion

BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation

Paper • 2205.13542 • Published May 26, 2022