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
QNN_CONTEXT_BINARY	float	qualcomm_qcs8550_proxy	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_qcs9075	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_sa7255p	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_sa8295p	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_sa8775p	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_snapdragon_8_elite_for_galaxy	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_snapdragon_8_elite_gen5	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_snapdragon_8gen3	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_snapdragon_x2_elite	QAIRT 2.43	Download
QNN_CONTEXT_BINARY	float	qualcomm_snapdragon_x_elite	QAIRT 2.43	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	5.732 ms	16 - 26 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	7.018 ms	24 - 24 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® X Elite	13.183 ms	23 - 23 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.854 ms	15 - 22 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8550 (Proxy)	12.629 ms	4 - 32 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	24.731 ms	17 - 20 MB	NPU
decoder	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.727 ms	12 - 19 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	5.717 ms	5 - 14 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	7.442 ms	5 - 5 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	13.171 ms	5 - 5 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	9.851 ms	5 - 12 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275 (Proxy)	92.849 ms	1 - 10 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	12.619 ms	5 - 6 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	24.486 ms	5 - 16 MB	NPU
decoder	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	7.714 ms	5 - 14 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	349.533 ms	51 - 60 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	380.439 ms	102 - 102 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® X Elite	689.856 ms	101 - 101 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	522.88 ms	32 - 44 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8550 (Proxy)	726.26 ms	0 - 116 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	705.74 ms	12 - 27 MB	NPU
encoder_1	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	422.574 ms	38 - 49 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	346.35 ms	18 - 28 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	382.894 ms	12 - 12 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	687.321 ms	12 - 12 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	522.601 ms	13 - 19 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275 (Proxy)	1100.241 ms	0 - 8 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	717.591 ms	13 - 14 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	704.883 ms	12 - 43 MB	NPU
encoder_1	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	421.049 ms	12 - 25 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2174.519 ms	351 - 361 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	2362.007 ms	1056 - 1056 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® X Elite	3434.434 ms	1058 - 1058 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	2665.392 ms	588 - 594 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8550 (Proxy)	3679.384 ms	593 - 606 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	3598.547 ms	233 - 236 MB	NPU
encoder_2	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	2449.521 ms	515 - 522 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	2195.452 ms	17 - 26 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	2398.265 ms	17 - 17 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	3539.157 ms	17 - 17 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	2739.151 ms	17 - 27 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275 (Proxy)	5606.428 ms	17 - 26 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	3547.128 ms	17 - 18 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	3505.941 ms	17 - 644 MB	NPU
encoder_2	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	2433.628 ms	17 - 29 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.766 ms	31 - 41 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® X2 Elite	7.78 ms	14 - 14 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® X Elite	12.219 ms	19 - 19 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.161 ms	30 - 36 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS8550 (Proxy)	12.403 ms	23 - 27 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Qualcomm® QCS9075	14.973 ms	18 - 39 MB	NPU
encoder_3	PRECOMPILED_QNN_ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.702 ms	13 - 20 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite Gen 5 Mobile	6.756 ms	18 - 29 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® X2 Elite	8.26 ms	19 - 19 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® X Elite	12.075 ms	19 - 19 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Gen 3 Mobile	9.228 ms	18 - 25 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8275 (Proxy)	25.294 ms	18 - 27 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS8550 (Proxy)	12.569 ms	19 - 20 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Qualcomm® QCS9075	15.062 ms	18 - 43 MB	NPU
encoder_3	QNN_CONTEXT_BINARY	float	Snapdragon® 8 Elite For Galaxy Mobile	7.789 ms	14 - 23 MB	NPU

License

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

References

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