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 1.24.1	Download
ONNX	w8a16_mixed_fp16	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.17.0	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	183.019 ms	20 - 20 MB	NPU
CVT	ONNX	float	Snapdragon® X Elite	288.642 ms	18 - 18 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Gen 3 Mobile	227.053 ms	8 - 2577 MB	NPU
CVT	ONNX	float	Qualcomm® QCS8550 (Proxy)	331.882 ms	0 - 24 MB	NPU
CVT	ONNX	float	Qualcomm® QCS9075	327.526 ms	7 - 10 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	190.833 ms	0 - 2218 MB	NPU
CVT	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	183.548 ms	8 - 2259 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® X2 Elite	223.535 ms	17 - 17 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® X Elite	364.569 ms	16 - 16 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	308.333 ms	15 - 1611 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite For Galaxy Mobile	263.664 ms	14 - 1174 MB	NPU
CVT	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite Gen 5 Mobile	217.599 ms	16 - 1268 MB	NPU
CVT	QNN_DLC	float	Snapdragon® X2 Elite	181.045 ms	7 - 7 MB	NPU
CVT	QNN_DLC	float	Snapdragon® X Elite	280.327 ms	7 - 7 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	232.914 ms	5 - 2551 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	420.107 ms	1 - 2187 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	321.73 ms	8 - 11 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8775P	291.527 ms	1 - 2187 MB	NPU
CVT	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	448.203 ms	5 - 2647 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA7255P	420.107 ms	1 - 2187 MB	NPU
CVT	QNN_DLC	float	Qualcomm® SA8295P	332.897 ms	0 - 2240 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	183.082 ms	7 - 2172 MB	NPU
CVT	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	167.215 ms	7 - 2258 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	225.117 ms	0 - 2557 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8275 (Proxy)	426.739 ms	0 - 2194 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8550 (Proxy)	335.078 ms	0 - 4 MB	NPU
CVT	TFLITE	float	Qualcomm® SA8775P	296.259 ms	0 - 2193 MB	NPU
CVT	TFLITE	float	Qualcomm® QCS8450 (Proxy)	465.477 ms	0 - 2733 MB	NPU
CVT	TFLITE	float	Qualcomm® SA7255P	426.739 ms	0 - 2194 MB	NPU
CVT	TFLITE	float	Qualcomm® SA8295P	339.762 ms	0 - 2290 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	186.793 ms	0 - 2235 MB	NPU
CVT	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	171.033 ms	0 - 2286 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.