FFNet-54S: Optimized for Qualcomm Devices

FFNet-54S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.

This is based on the implementation of FFNet-54S 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	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.17.0	Download

For more device-specific assets and performance metrics, visit FFNet-54S 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 FFNet-54S on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: ffnet54S_dBBB_cityscapes_state_dict_quarts
• Input resolution: 2048x1024
• Number of output classes: 19
• Number of parameters: 18.0M
• Model size (float): 68.8 MB
• Model size (w8a8): 17.5 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-54S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	13.443 ms	29 - 256 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® X2 Elite	14.948 ms	22 - 22 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® X Elite	34.004 ms	24 - 24 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	24.125 ms	30 - 315 MB	NPU
FFNet-54S	ONNX	float	Qualcomm® QCS8550 (Proxy)	34.307 ms	24 - 27 MB	NPU
FFNet-54S	ONNX	float	Qualcomm® QCS9075	52.365 ms	24 - 51 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	18.079 ms	5 - 202 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	6.922 ms	7 - 210 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® X2 Elite	7.392 ms	13 - 13 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® X Elite	11.157 ms	12 - 12 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	6.947 ms	7 - 264 MB	NPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCS6490	408.007 ms	183 - 239 MB	CPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	10.549 ms	0 - 16 MB	NPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCS9075	12.875 ms	6 - 9 MB	NPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCM6690	433.736 ms	153 - 162 MB	CPU
FFNet-54S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	11.622 ms	1 - 200 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	429.655 ms	106 - 115 MB	CPU
FFNet-54S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	13.598 ms	8 - 252 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® X2 Elite	15.857 ms	24 - 24 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® X Elite	39.583 ms	24 - 24 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	26.731 ms	11 - 302 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	154.549 ms	24 - 226 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	38.137 ms	24 - 359 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® SA8775P	53.589 ms	24 - 224 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS9075	66.232 ms	24 - 52 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	77.665 ms	6 - 287 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® SA7255P	154.549 ms	24 - 226 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® SA8295P	58.607 ms	24 - 221 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	19.421 ms	16 - 237 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.484 ms	6 - 244 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® X2 Elite	6.509 ms	6 - 6 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® X Elite	16.716 ms	6 - 6 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.139 ms	6 - 263 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS6490	71.605 ms	4 - 12 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	35.411 ms	6 - 205 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	15.867 ms	6 - 46 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® SA8775P	16.337 ms	6 - 206 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS9075	18.772 ms	6 - 14 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCM6690	142.725 ms	6 - 241 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	19.966 ms	6 - 265 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® SA7255P	35.411 ms	6 - 205 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® SA8295P	21.462 ms	6 - 208 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.588 ms	6 - 221 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	20.356 ms	6 - 217 MB	NPU
FFNet-54S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	13.612 ms	2 - 265 MB	NPU
FFNet-54S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	26.455 ms	2 - 345 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	154.489 ms	3 - 228 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	38.118 ms	2 - 5 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® SA8775P	53.622 ms	2 - 226 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	78.23 ms	2 - 335 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® SA7255P	154.489 ms	3 - 228 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® SA8295P	58.624 ms	2 - 225 MB	NPU
FFNet-54S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	19.426 ms	1 - 243 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.205 ms	1 - 237 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	5.888 ms	0 - 261 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS6490	55.99 ms	1 - 27 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	22.847 ms	1 - 199 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	8.26 ms	1 - 3 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® SA8775P	8.89 ms	1 - 200 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS9075	10.034 ms	0 - 27 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCM6690	117.615 ms	1 - 238 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	13.15 ms	1 - 262 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® SA7255P	22.847 ms	1 - 199 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® SA8295P	13.117 ms	1 - 202 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.444 ms	0 - 216 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	12.697 ms	1 - 218 MB	NPU

License

• The license for the original implementation of FFNet-54S can be found here.

References

• Simple and Efficient Architectures for 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.