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® X2 Elite	18.102 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® X Elite	37.773 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	26.986 ms	0 - 302 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS8550 (Proxy)	38.05 ms	24 - 27 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS9075	59.191 ms	24 - 51 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	20.318 ms	5 - 214 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	17.168 ms	30 - 259 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X2 Elite	7.934 ms	22 - 22 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X Elite	14.86 ms	21 - 21 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	11.575 ms	7 - 297 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS6490	496.283 ms	168 - 221 MB	CPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	15.576 ms	0 - 24 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS9075	14.497 ms	6 - 9 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCM6690	536.759 ms	136 - 146 MB	CPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.773 ms	1 - 209 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	534.017 ms	145 - 156 MB	CPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.308 ms	0 - 219 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X2 Elite	18.007 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X Elite	43.789 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	29.414 ms	76 - 389 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	186.955 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	42.249 ms	24 - 28 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8775P	60.75 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS9075	73.12 ms	26 - 54 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	84.436 ms	3 - 298 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA7255P	186.955 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8295P	66.023 ms	24 - 230 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	21.466 ms	17 - 245 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	15.524 ms	8 - 258 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X2 Elite	7.039 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X Elite	17.701 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.644 ms	6 - 287 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS6490	72.774 ms	5 - 13 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	39.127 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	16.72 ms	6 - 42 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8775P	77.742 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS9075	20.341 ms	6 - 14 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCM6690	163.405 ms	6 - 253 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	24.122 ms	6 - 285 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA7255P	39.127 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8295P	23.343 ms	6 - 219 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.936 ms	6 - 234 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	22.335 ms	6 - 237 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.885 ms	6 - 256 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	29.374 ms	1 - 388 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	186.875 ms	0 - 245 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	42.398 ms	3 - 5 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8775P	60.867 ms	2 - 247 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS9075	72.983 ms	0 - 82 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	85.435 ms	2 - 375 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA7255P	186.875 ms	0 - 245 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8295P	66.052 ms	2 - 244 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	21.647 ms	0 - 259 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	15.398 ms	1 - 284 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	6.451 ms	1 - 285 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS6490	57.155 ms	1 - 36 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	26.382 ms	1 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	9.074 ms	1 - 3 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8775P	9.771 ms	0 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS9075	10.967 ms	1 - 36 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCM6690	138.255 ms	1 - 246 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	18.372 ms	1 - 286 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA7255P	26.382 ms	1 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8295P	14.992 ms	1 - 214 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.954 ms	0 - 230 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	14.678 ms	0 - 230 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.594 ms	0 - 248 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.