FFNet-40S: Optimized for Qualcomm Devices

FFNet-40S 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-40S 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 2.37, ONNX Runtime 1.23.0	Download
ONNX	w8a8	Universal	QAIRT 2.37, ONNX Runtime 1.23.0	Download
QNN_DLC	float	Universal	QAIRT 2.42	Download
QNN_DLC	w8a8	Universal	QAIRT 2.42	Download
TFLITE	float	Universal	QAIRT 2.42, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.42, TFLite 2.17.0	Download

For more device-specific assets and performance metrics, visit FFNet-40S 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-40S on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: ffnet40S_dBBB_cityscapes_state_dict_quarts
• Input resolution: 2048x1024
• Number of output classes: 19
• Number of parameters: 13.9M
• Model size (float): 53.1 MB
• Model size (w8a8): 13.5 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-40S	ONNX	float	Snapdragon® X Elite	28.126 ms	24 - 24 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	19.18 ms	0 - 214 MB	NPU
FFNet-40S	ONNX	float	Qualcomm® QCS8550 (Proxy)	28.635 ms	24 - 55 MB	NPU
FFNet-40S	ONNX	float	Qualcomm® QCS9075	48.062 ms	24 - 27 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	14.112 ms	8 - 154 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	10.615 ms	31 - 194 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® X Elite	10.29 ms	8 - 8 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	6.795 ms	7 - 197 MB	NPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCS6490	360.996 ms	200 - 239 MB	CPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	9.811 ms	0 - 33 MB	NPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCS9075	13.756 ms	6 - 9 MB	NPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCM6690	369.284 ms	213 - 222 MB	CPU
FFNet-40S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	6.801 ms	1 - 141 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	360.344 ms	126 - 135 MB	CPU
FFNet-40S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.847 ms	0 - 145 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® X Elite	37.623 ms	24 - 24 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	24.903 ms	20 - 302 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	135.812 ms	24 - 220 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	36.334 ms	24 - 56 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® SA8775P	227.588 ms	24 - 223 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS9075	61.786 ms	24 - 52 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	115.457 ms	22 - 305 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® SA7255P	135.812 ms	24 - 220 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® SA8295P	64.029 ms	24 - 227 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	18.206 ms	14 - 239 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	12.618 ms	10 - 248 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® X Elite	15.965 ms	6 - 6 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	10.574 ms	6 - 248 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS6490	66.653 ms	6 - 14 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	33.037 ms	6 - 198 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	15.024 ms	6 - 8 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® SA8775P	15.768 ms	6 - 198 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS9075	18.747 ms	8 - 16 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCM6690	131.7 ms	6 - 238 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	21.633 ms	6 - 249 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® SA7255P	33.037 ms	6 - 198 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® SA8295P	20.252 ms	6 - 201 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.246 ms	6 - 215 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	19.063 ms	6 - 220 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.228 ms	6 - 236 MB	NPU
FFNet-40S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	25.114 ms	2 - 321 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	135.962 ms	2 - 214 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	36.322 ms	2 - 5 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® SA8775P	226.95 ms	3 - 214 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS9075	61.957 ms	0 - 56 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	115.674 ms	2 - 325 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® SA7255P	135.962 ms	2 - 214 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® SA8295P	64.05 ms	2 - 222 MB	NPU
FFNet-40S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	18.109 ms	1 - 237 MB	NPU
FFNet-40S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	12.596 ms	0 - 251 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	5.448 ms	1 - 246 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS6490	52.113 ms	1 - 23 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	20.37 ms	0 - 190 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	7.599 ms	1 - 3 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® SA8775P	8.225 ms	1 - 192 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS9075	9.523 ms	0 - 22 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCM6690	102.975 ms	1 - 231 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	12.19 ms	1 - 244 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® SA7255P	20.37 ms	0 - 190 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® SA8295P	11.818 ms	1 - 196 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.029 ms	0 - 207 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	11.445 ms	0 - 211 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	2.949 ms	1 - 230 MB	NPU

License

• The license for the original implementation of FFNet-40S 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.