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® X Elite	37.916 ms	30 - 30 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	27.087 ms	2 - 303 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS8550 (Proxy)	38.819 ms	24 - 55 MB	NPU
FFNet-78S	ONNX	float	Qualcomm® QCS9075	59.465 ms	24 - 51 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	20.025 ms	7 - 210 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	17.186 ms	29 - 258 MB	NPU
FFNet-78S	ONNX	float	Snapdragon® X2 Elite	18.088 ms	30 - 30 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	11.802 ms	7 - 294 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS6490	488.89 ms	168 - 224 MB	CPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	15.459 ms	0 - 392 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCS9075	14.458 ms	6 - 9 MB	NPU
FFNet-78S	ONNX	w8a8	Qualcomm® QCM6690	536.573 ms	132 - 142 MB	CPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.799 ms	1 - 214 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	530.544 ms	147 - 157 MB	CPU
FFNet-78S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.305 ms	2 - 220 MB	NPU
FFNet-78S	ONNX	w8a8	Snapdragon® X2 Elite	7.905 ms	22 - 22 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X Elite	43.796 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	29.532 ms	24 - 332 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	186.934 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	42.346 ms	24 - 39 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8775P	60.781 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS9075	73.761 ms	24 - 52 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	84.171 ms	2 - 292 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA7255P	186.934 ms	24 - 235 MB	NPU
FFNet-78S	QNN_DLC	float	Qualcomm® SA8295P	66.03 ms	24 - 230 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	21.693 ms	22 - 250 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	15.51 ms	24 - 276 MB	NPU
FFNet-78S	QNN_DLC	float	Snapdragon® X2 Elite	18.002 ms	24 - 24 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X Elite	17.748 ms	6 - 6 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.653 ms	6 - 285 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS6490	72.404 ms	6 - 14 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	39.165 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	16.791 ms	6 - 8 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8775P	77.741 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS9075	19.433 ms	6 - 14 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCM6690	165.866 ms	6 - 253 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	24.066 ms	6 - 288 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA7255P	39.165 ms	6 - 217 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Qualcomm® SA8295P	23.322 ms	6 - 220 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.965 ms	6 - 233 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	22.462 ms	6 - 237 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.897 ms	6 - 255 MB	NPU
FFNet-78S	QNN_DLC	w8a8	Snapdragon® X2 Elite	7.054 ms	6 - 6 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	29.472 ms	0 - 387 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	186.973 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	43.022 ms	2 - 5 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8775P	282.888 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS9075	72.981 ms	0 - 82 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	84.408 ms	1 - 374 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA7255P	186.973 ms	3 - 248 MB	NPU
FFNet-78S	TFLITE	float	Qualcomm® SA8295P	66.076 ms	2 - 244 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	21.685 ms	1 - 265 MB	NPU
FFNet-78S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	15.447 ms	2 - 284 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	6.444 ms	1 - 284 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS6490	57.085 ms	1 - 36 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	26.379 ms	1 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	9.065 ms	1 - 3 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8775P	42.19 ms	1 - 211 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS9075	10.987 ms	1 - 36 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCM6690	140.078 ms	1 - 247 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	18.364 ms	4 - 289 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA7255P	26.379 ms	1 - 210 MB	NPU
FFNet-78S	TFLITE	w8a8	Qualcomm® SA8295P	14.995 ms	1 - 214 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.956 ms	0 - 229 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	14.715 ms	0 - 230 MB	NPU
FFNet-78S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.594 ms	1 - 249 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.