CVT: Optimized for Qualcomm Devices

Cross-View Transformer generates real-time bird's-eye view maps from multiple vehicle cameras for autonomous driving.

This is based on the implementation of CVT 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 2.42, ONNX Runtime 1.25.0	Download
ONNX	w8a16_mixed_fp16	Universal	QAIRT 2.42, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

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

Model Details

Model Type: Model_use_case.driver_assistance

Model Stats:

• Model checkpoint: vehicles_50k.pt
• Inference latency: RealTime
• Input resolution: 1x6x3x224x480
• Number of parameters: 1.33M
• Model size (float): 5.18 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
CVT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	180.758 ms	8 - 2258 MB	NPU
CVT	ONNX	float	Snapdragon® X2 Elite	182.295 ms	205 - 205 MB	NPU
CVT	ONNX	float	Snapdragon® X Elite	284.357 ms	147 - 147 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	243.389 ms	8 - 2601 MB	NPU
CVT	ONNX	float	Qualcomm® QCS8550 (Proxy)	306.931 ms	0 - 25 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	186.596 ms	5 - 2270 MB	NPU
CVT	ONNX	float	Qualcomm® QCS9075	321.296 ms	7 - 52 MB	NPU
CVT	ONNX	float	Qualcomm® QCS8750	186.596 ms	5 - 2270 MB	NPU
CVT	ONNX	float	Qualcomm® QCS7181	284.357 ms	147 - 147 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite Gen 5 Mobile	218.945 ms	15 - 1246 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	298.525 ms	15 - 1617 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite For Galaxy Mobile	261.974 ms	13 - 1107 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Qualcomm® QCS8750	261.974 ms	13 - 1107 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	168.835 ms	7 - 2297 MB	NPU
CVT	QNN_DLC	float	Snapdragon® X2 Elite	182.659 ms	7 - 7 MB	NPU
CVT	QNN_DLC	float	Snapdragon® X Elite	286.587 ms	7 - 7 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	222.542 ms	0 - 2569 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8275	425.777 ms	2 - 2214 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	333.185 ms	8 - 12 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8775P	293.056 ms	2 - 2213 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8650P	293.056 ms	2 - 2213 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8255P	293.056 ms	2 - 2213 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	441.965 ms	7 - 2719 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA7255P	425.777 ms	2 - 2214 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8295P	334.17 ms	0 - 2323 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	181.922 ms	0 - 2290 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS9075	322.616 ms	7 - 17 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8750	181.922 ms	0 - 2290 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS7181	286.587 ms	7 - 7 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	166.616 ms	0 - 2305 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	224.577 ms	0 - 2670 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8550 (Proxy)	321.143 ms	0 - 4 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8450 (Proxy)	440.053 ms	1 - 2726 MB	NPU
CVT	TFLITE	float	Qualcomm® SA8295P	330.943 ms	0 - 2306 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	180.728 ms	1 - 2364 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS9075	318.129 ms	0 - 36 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8750	180.728 ms	1 - 2364 MB	NPU

License

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

References

• Cross-view Transformers for real-time Map-view Semantic Segmentation
• 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.