FFNet-78S: Optimized for Qualcomm Devices

FFNet-78S 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-78S 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-78S 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-78S on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: ffnet78S_dBBB_cityscapes_state_dict_quarts
• Input resolution: 2048x1024
• Number of output classes: 19
• Number of parameters: 27.5M
• Model size (float): 105 MB
• Model size (w8a8): 26.7 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-78S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	17.26 ms	29 - 257 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® X2 Elite	18.053 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® X Elite	37.729 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	27.076 ms	0 - 303 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS8550 (Proxy)	38.291 ms	24 - 58 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS9075	59.563 ms	24 - 51 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	20.275 ms	7 - 210 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.298 ms	0 - 219 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X2 Elite	7.882 ms	22 - 22 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X Elite	14.824 ms	21 - 21 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	11.625 ms	7 - 294 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS6490	487.641 ms	162 - 216 MB	CPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	15.459 ms	0 - 25 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS9075	14.549 ms	6 - 9 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCM6690	535.901 ms	176 - 186 MB	CPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.771 ms	1 - 212 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	535.24 ms	144 - 155 MB	CPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	15.517 ms	8 - 258 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X2 Elite	48.916 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X Elite	43.814 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	29.384 ms	8 - 315 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	187.008 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	42.303 ms	24 - 26 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8775P	60.742 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS9075	72.995 ms	24 - 52 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	83.966 ms	2 - 295 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA7255P	187.008 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8295P	65.995 ms	24 - 229 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	21.718 ms	16 - 244 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.886 ms	6 - 256 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X2 Elite	7.029 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X Elite	17.682 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.636 ms	6 - 287 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS6490	72.359 ms	4 - 12 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	39.144 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	16.769 ms	6 - 8 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8775P	17.199 ms	6 - 218 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS9075	19.955 ms	8 - 16 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCM6690	142.792 ms	6 - 254 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	24.066 ms	6 - 286 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA7255P	39.144 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8295P	23.34 ms	6 - 220 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.961 ms	6 - 233 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	22.418 ms	6 - 237 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	15.385 ms	1 - 286 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	29.497 ms	2 - 394 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	186.902 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	42.742 ms	2 - 5 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8775P	60.834 ms	0 - 245 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	86.014 ms	0 - 371 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA7255P	186.902 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8295P	66.02 ms	2 - 244 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	21.58 ms	1 - 264 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.603 ms	1 - 249 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	6.441 ms	0 - 285 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS6490	57.031 ms	1 - 36 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	26.392 ms	1 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	9.049 ms	0 - 2 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8775P	42.159 ms	1 - 211 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS9075	10.973 ms	0 - 35 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCM6690	124.774 ms	1 - 248 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	18.303 ms	1 - 285 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA7255P	26.392 ms	1 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8295P	14.994 ms	1 - 214 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.958 ms	0 - 229 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	14.611 ms	0 - 230 MB	NPU

License

• The license for the original implementation of FFNet-78S 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.