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 1.24.3	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.19.1	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.19.1	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	17.137 ms	27 - 257 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® X2 Elite	17.864 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® X Elite	37.616 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	27.249 ms	1 - 303 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS8550 (Proxy)	38.207 ms	24 - 30 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS9075	59.612 ms	24 - 50 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	20.356 ms	7 - 209 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.293 ms	2 - 221 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X2 Elite	7.923 ms	22 - 22 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X Elite	14.864 ms	21 - 21 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	11.668 ms	7 - 297 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS6490	492.171 ms	162 - 228 MB	CPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	15.574 ms	0 - 24 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS9075	14.501 ms	6 - 9 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCM6690	532.971 ms	125 - 135 MB	CPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.772 ms	1 - 209 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	534.544 ms	147 - 157 MB	CPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	15.493 ms	11 - 263 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X2 Elite	17.997 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X Elite	43.811 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	29.282 ms	24 - 333 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	186.927 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	42.282 ms	24 - 403 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8775P	60.766 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS9075	73.124 ms	24 - 52 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	84.572 ms	2 - 296 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA7255P	186.927 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8295P	66.021 ms	24 - 230 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	21.644 ms	17 - 247 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.887 ms	6 - 256 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X2 Elite	14.045 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X Elite	17.722 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.665 ms	6 - 285 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS6490	72.458 ms	8 - 16 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	39.132 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	16.805 ms	6 - 109 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8775P	17.192 ms	6 - 218 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS9075	20.092 ms	6 - 14 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCM6690	163.774 ms	6 - 253 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	24.37 ms	6 - 287 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA7255P	39.132 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8295P	23.367 ms	6 - 220 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.939 ms	6 - 233 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	22.306 ms	6 - 237 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	15.526 ms	2 - 285 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	29.591 ms	1 - 393 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	186.958 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	42.738 ms	2 - 5 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8775P	60.877 ms	2 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS9075	72.979 ms	0 - 82 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	85.307 ms	3 - 374 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA7255P	186.958 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8295P	66.105 ms	0 - 241 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	21.617 ms	2 - 265 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.606 ms	1 - 248 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	6.453 ms	1 - 286 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS6490	57.316 ms	0 - 36 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	26.368 ms	0 - 209 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	9.043 ms	0 - 3 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8775P	9.761 ms	1 - 212 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS9075	10.988 ms	0 - 35 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCM6690	140.957 ms	1 - 248 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	18.428 ms	0 - 283 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA7255P	26.368 ms	0 - 209 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8295P	14.999 ms	1 - 214 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.978 ms	0 - 229 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	14.631 ms	1 - 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.