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 2.42, 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-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® X2 Elite	13.263 ms	22 - 22 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® X Elite	31.492 ms	24 - 24 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	22.483 ms	30 - 303 MB	NPU
FFNet-40S	ONNX	float	Qualcomm® QCS8550 (Proxy)	31.785 ms	24 - 26 MB	NPU
FFNet-40S	ONNX	float	Qualcomm® QCS9075	48.061 ms	24 - 27 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	16.024 ms	8 - 209 MB	NPU
FFNet-40S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	12.347 ms	26 - 247 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® X2 Elite	6.929 ms	9 - 9 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® X Elite	10.461 ms	8 - 8 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	6.568 ms	7 - 248 MB	NPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCS6490	360.076 ms	202 - 239 MB	CPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	9.833 ms	0 - 14 MB	NPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCS9075	12.946 ms	6 - 9 MB	NPU
FFNet-40S	ONNX	w8a8	Qualcomm® QCM6690	365.578 ms	215 - 224 MB	CPU
FFNet-40S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.726 ms	1 - 192 MB	NPU
FFNet-40S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	361.936 ms	164 - 173 MB	CPU
FFNet-40S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	6.563 ms	2 - 198 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® X2 Elite	14.557 ms	24 - 24 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® X Elite	37.107 ms	24 - 24 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	24.737 ms	24 - 303 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	134.465 ms	15 - 210 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	36.181 ms	24 - 200 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® SA8775P	48.805 ms	24 - 220 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS9075	62.122 ms	24 - 52 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	116.564 ms	23 - 307 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® SA7255P	134.465 ms	15 - 210 MB	NPU
FFNet-40S	QNN_DLC	float	Qualcomm® SA8295P	64.889 ms	24 - 226 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	18.149 ms	21 - 248 MB	NPU
FFNet-40S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	12.576 ms	24 - 262 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® X2 Elite	6.227 ms	6 - 6 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® X Elite	15.979 ms	6 - 6 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	10.571 ms	6 - 249 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS6490	67.566 ms	4 - 12 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	32.881 ms	6 - 198 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	15.145 ms	6 - 8 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® SA8775P	15.757 ms	6 - 199 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS9075	18.22 ms	6 - 14 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCM6690	127.392 ms	6 - 239 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	21.509 ms	6 - 250 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® SA7255P	32.881 ms	6 - 198 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Qualcomm® SA8295P	20.179 ms	6 - 201 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.174 ms	6 - 213 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	18.969 ms	6 - 220 MB	NPU
FFNet-40S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.211 ms	6 - 236 MB	NPU
FFNet-40S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	24.885 ms	1 - 317 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	134.647 ms	3 - 213 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	36.505 ms	2 - 5 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® SA8775P	48.866 ms	2 - 213 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS9075	61.819 ms	0 - 56 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	115.735 ms	2 - 324 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® SA7255P	134.647 ms	3 - 213 MB	NPU
FFNet-40S	TFLITE	float	Qualcomm® SA8295P	64.956 ms	0 - 219 MB	NPU
FFNet-40S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	17.949 ms	1 - 240 MB	NPU
FFNet-40S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	12.609 ms	2 - 253 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	5.444 ms	1 - 248 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS6490	52.251 ms	1 - 23 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	20.416 ms	1 - 191 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	7.616 ms	1 - 2 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® SA8775P	8.228 ms	1 - 192 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS9075	9.533 ms	0 - 22 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCM6690	101.921 ms	1 - 232 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	12.264 ms	1 - 247 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® SA7255P	20.416 ms	1 - 191 MB	NPU
FFNet-40S	TFLITE	w8a8	Qualcomm® SA8295P	11.815 ms	1 - 195 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.047 ms	0 - 207 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	11.564 ms	0 - 213 MB	NPU
FFNet-40S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	2.941 ms	1 - 229 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.