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 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a16_mixed_fp16	Universal	QAIRT 2.45, 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® X2 Elite	183.908 ms	173 - 173 MB	NPU
CVT	ONNX	float	Snapdragon® X Elite	274.2 ms	142 - 142 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	245.125 ms	8 - 2683 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Gen 1 Mobile	424.309 ms	10 - 2821 MB	NPU
CVT	ONNX	float	Qualcomm® QCS8550 (Proxy)	325.517 ms	0 - 24 MB	NPU
CVT	ONNX	float	Qualcomm® QCS8450	424.309 ms	10 - 2821 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Elite Mobile	187.62 ms	3 - 2380 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	167.214 ms	0 - 2411 MB	NPU
CVT	ONNX	float	Qualcomm® QCS9075	321.618 ms	7 - 53 MB	NPU
CVT	ONNX	float	Qualcomm® QCS8750	187.62 ms	3 - 2380 MB	NPU
CVT	ONNX	float	Qualcomm® QCS7181	274.2 ms	142 - 142 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	253.311 ms	15 - 2643 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Qualcomm® QCS8550 (Proxy)	355.673 ms	13 - 18 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite Mobile	217.182 ms	12 - 2415 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite Gen 5 Mobile	198.04 ms	12 - 2507 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Qualcomm® QCS9075	425.412 ms	14 - 59 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Qualcomm® QCS8750	217.182 ms	12 - 2415 MB	NPU
CVT	QNN_DLC	float	Snapdragon® X2 Elite	182.322 ms	7 - 7 MB	NPU
CVT	QNN_DLC	float	Snapdragon® X Elite	286.885 ms	7 - 7 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	222.363 ms	7 - 2575 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	441.585 ms	7 - 2718 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8275	425.513 ms	2 - 2213 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	331.6 ms	8 - 13 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8450	441.585 ms	7 - 2718 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Elite Mobile	184.913 ms	0 - 2290 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8295P	332.666 ms	0 - 2323 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	168.158 ms	7 - 2319 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS9075	322.622 ms	7 - 17 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA7255P	425.513 ms	2 - 2213 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8750	184.913 ms	0 - 2290 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS7181	286.885 ms	7 - 7 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	215.76 ms	1 - 2673 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	416.47 ms	1 - 2724 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8550 (Proxy)	306.789 ms	0 - 3 MB	NPU
CVT	TFLITE	float	Qualcomm® SA8775P	572.418 ms	48 - 60 MB	CPU
CVT	TFLITE	float	Qualcomm® SA8650P	572.418 ms	48 - 60 MB	CPU
CVT	TFLITE	float	Qualcomm® SA8255P	572.418 ms	48 - 60 MB	CPU
CVT	TFLITE	float	Qualcomm® QCS8450	416.47 ms	1 - 2724 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Elite Mobile	182.607 ms	0 - 2363 MB	NPU
CVT	TFLITE	float	Qualcomm® SA8295P	332.722 ms	0 - 2306 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	167.605 ms	0 - 2315 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS9075	317.919 ms	0 - 36 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8750	182.607 ms	0 - 2363 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.