FFNet-54S: Optimized for Qualcomm Devices

FFNet-54S 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-54S 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-54S 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-54S on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: ffnet54S_dBBB_cityscapes_state_dict_quarts
• Input resolution: 2048x1024
• Number of output classes: 19
• Number of parameters: 18.0M
• Model size (float): 68.8 MB
• Model size (w8a8): 17.5 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-54S	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	13.502 ms	31 - 261 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® X2 Elite	14.938 ms	22 - 22 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® X Elite	34.028 ms	24 - 24 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® 8 Gen 3 Mobile	24.363 ms	30 - 317 MB	NPU
FFNet-54S	ONNX	float	Qualcomm® QCS8550 (Proxy)	33.806 ms	24 - 27 MB	NPU
FFNet-54S	ONNX	float	Qualcomm® QCS9075	52.797 ms	24 - 51 MB	NPU
FFNet-54S	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	18.15 ms	5 - 202 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	6.916 ms	7 - 210 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® X2 Elite	7.401 ms	13 - 13 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® X Elite	11.147 ms	12 - 12 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	6.957 ms	7 - 265 MB	NPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCS6490	422.119 ms	183 - 239 MB	CPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	10.575 ms	0 - 16 MB	NPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCS9075	12.881 ms	6 - 9 MB	NPU
FFNet-54S	ONNX	w8a8	Qualcomm® QCM6690	431.658 ms	166 - 175 MB	CPU
FFNet-54S	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	11.672 ms	1 - 200 MB	NPU
FFNet-54S	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	430.164 ms	180 - 190 MB	CPU
FFNet-54S	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	13.696 ms	24 - 267 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® X2 Elite	15.917 ms	24 - 24 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® X Elite	39.458 ms	24 - 24 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	26.64 ms	24 - 316 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	154.393 ms	24 - 226 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	37.881 ms	24 - 26 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® SA8775P	53.636 ms	24 - 225 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS9075	66.302 ms	24 - 52 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	77.457 ms	4 - 284 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® SA7255P	154.393 ms	24 - 226 MB	NPU
FFNet-54S	QNN_DLC	float	Qualcomm® SA8295P	58.633 ms	24 - 221 MB	NPU
FFNet-54S	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	19.404 ms	13 - 235 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	5.483 ms	6 - 243 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® X2 Elite	6.476 ms	6 - 6 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® X Elite	16.693 ms	6 - 6 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.191 ms	6 - 264 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS6490	72.034 ms	5 - 13 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	35.414 ms	6 - 205 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	15.858 ms	6 - 8 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® SA8775P	16.354 ms	6 - 206 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS9075	18.83 ms	8 - 16 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCM6690	142.263 ms	6 - 242 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	20.076 ms	6 - 263 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® SA7255P	35.414 ms	6 - 205 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Qualcomm® SA8295P	21.445 ms	6 - 207 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	7.597 ms	6 - 221 MB	NPU
FFNet-54S	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	20.832 ms	6 - 217 MB	NPU
FFNet-54S	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	13.649 ms	2 - 264 MB	NPU
FFNet-54S	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	27.002 ms	1 - 345 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS8275 (Proxy)	154.634 ms	3 - 227 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS8550 (Proxy)	38.607 ms	2 - 5 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® SA8775P	53.577 ms	2 - 226 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS9075	66.845 ms	0 - 64 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® QCS8450 (Proxy)	77.919 ms	3 - 336 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® SA7255P	154.634 ms	3 - 227 MB	NPU
FFNet-54S	TFLITE	float	Qualcomm® SA8295P	58.65 ms	2 - 225 MB	NPU
FFNet-54S	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	19.534 ms	1 - 247 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3.189 ms	1 - 236 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	5.88 ms	1 - 261 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS6490	55.613 ms	1 - 27 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	22.871 ms	1 - 198 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	8.267 ms	1 - 40 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® SA8775P	8.87 ms	1 - 200 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS9075	10.044 ms	0 - 26 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCM6690	116.677 ms	1 - 238 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	13.1 ms	0 - 260 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® SA7255P	22.871 ms	1 - 198 MB	NPU
FFNet-54S	TFLITE	w8a8	Qualcomm® SA8295P	13.158 ms	1 - 201 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	4.431 ms	0 - 215 MB	NPU
FFNet-54S	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	12.655 ms	1 - 218 MB	NPU

License

• The license for the original implementation of FFNet-54S 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.