FFNet-122NS-LowRes: Optimized for Qualcomm Devices

FFNet-122NS-LowRes 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-122NS-LowRes 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 2.37, ONNX Runtime 1.23.0	Download
ONNX	w8a8	Universal	QAIRT 2.37, ONNX Runtime 1.23.0	Download
QNN_DLC	float	Universal	QAIRT 2.42	Download
QNN_DLC	w8a8	Universal	QAIRT 2.42	Download
TFLITE	float	Universal	QAIRT 2.42, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.42, TFLite 2.17.0	Download

For more device-specific assets and performance metrics, visit FFNet-122NS-LowRes 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-122NS-LowRes on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: ffnet122NS_CCC_cityscapes_state_dict_quarts_pre_down
• Input resolution: 1024x512
• Number of output classes: 19
• Number of parameters: 32.1M
• Model size (float): 123 MB
• Model size (w8a8): 31.3 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-122NS-LowRes	ONNX	float	Snapdragon® X Elite	7.035 ms	56 - 56 MB	NPU
FFNet-122NS-LowRes	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.043 ms	3 - 166 MB	NPU
FFNet-122NS-LowRes	ONNX	float	Qualcomm® QCS8550 (Proxy)	7.045 ms	0 - 59 MB	NPU
FFNet-122NS-LowRes	ONNX	float	Qualcomm® QCS9075	10.371 ms	6 - 15 MB	NPU
FFNet-122NS-LowRes	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	6.258 ms	3 - 126 MB	NPU
FFNet-122NS-LowRes	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.671 ms	0 - 125 MB	NPU
FFNet-122NS-LowRes	ONNX	w8a8	Snapdragon® X Elite	2.793 ms	30 - 30 MB	NPU
FFNet-122NS-LowRes	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	2.036 ms	0 - 220 MB	NPU
FFNet-122NS-LowRes	ONNX	w8a8	Qualcomm® QCS6490	91.933 ms	54 - 153 MB	CPU
FFNet-122NS-LowRes	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	2.86 ms	0 - 37 MB	NPU
FFNet-122NS-LowRes	ONNX	w8a8	Qualcomm® QCS9075	3.305 ms	1 - 4 MB	NPU
FFNet-122NS-LowRes	ONNX	w8a8	Qualcomm® QCM6690	84.154 ms	60 - 73 MB	CPU
FFNet-122NS-LowRes	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	1.716 ms	0 - 157 MB	NPU
FFNet-122NS-LowRes	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	83.743 ms	61 - 75 MB	CPU
FFNet-122NS-LowRes	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.515 ms	0 - 157 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Snapdragon® X Elite	13.589 ms	6 - 6 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	8.817 ms	0 - 201 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	39.547 ms	0 - 160 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	13.343 ms	6 - 9 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® SA8775P	16.002 ms	0 - 162 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® QCS9075	16.888 ms	6 - 14 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	29.955 ms	6 - 197 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® SA7255P	39.547 ms	0 - 160 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Qualcomm® SA8295P	17.669 ms	0 - 152 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	6.99 ms	0 - 163 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.654 ms	6 - 172 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Snapdragon® X Elite	4.839 ms	2 - 2 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.191 ms	2 - 130 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® QCS6490	13.871 ms	4 - 7 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	9.288 ms	1 - 76 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.5 ms	2 - 3 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® SA8775P	21.084 ms	1 - 76 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® QCS9075	5.871 ms	2 - 5 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® QCM6690	27.238 ms	2 - 199 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	6.928 ms	2 - 126 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® SA7255P	9.288 ms	1 - 76 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Qualcomm® SA8295P	6.037 ms	2 - 74 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.194 ms	2 - 79 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	5.687 ms	2 - 193 MB	NPU
FFNet-122NS-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.803 ms	2 - 80 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	8.64 ms	1 - 280 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® QCS8275 (Proxy)	39.759 ms	1 - 194 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® QCS8550 (Proxy)	12.666 ms	1 - 3 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® SA8775P	16.045 ms	1 - 194 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® QCS9075	16.8 ms	0 - 70 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® QCS8450 (Proxy)	30.193 ms	1 - 271 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® SA7255P	39.759 ms	1 - 194 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Qualcomm® SA8295P	17.656 ms	1 - 191 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	6.882 ms	0 - 199 MB	NPU
FFNet-122NS-LowRes	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	5.061 ms	1 - 199 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	1.872 ms	0 - 136 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® QCS6490	9.208 ms	0 - 35 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	5.961 ms	0 - 71 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	2.619 ms	0 - 35 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® SA8775P	3.086 ms	0 - 74 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® QCS9075	3.041 ms	0 - 35 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® QCM6690	23.267 ms	0 - 198 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	3.285 ms	0 - 129 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® SA7255P	5.961 ms	0 - 71 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Qualcomm® SA8295P	3.868 ms	0 - 69 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	1.453 ms	0 - 69 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	3.844 ms	0 - 189 MB	NPU
FFNet-122NS-LowRes	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.304 ms	0 - 76 MB	NPU

License

• The license for the original implementation of FFNet-122NS-LowRes 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.