FFNet-78S-LowRes: Optimized for Qualcomm Devices

FFNet-78S-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-78S-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-78S-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-78S-LowRes on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: ffnet78S_BCC_cityscapes_state_dict_quarts_pre_down
• Input resolution: 1024x512
• Number of output classes: 19
• Number of parameters: 26.8M
• Model size (float): 102 MB
• Model size (w8a8): 26.0 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-78S-LowRes	ONNX	float	Snapdragon® X Elite	8.252 ms	47 - 47 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.792 ms	3 - 157 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® QCS8550 (Proxy)	7.989 ms	0 - 82 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® QCS9075	13.167 ms	6 - 15 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	6.927 ms	2 - 123 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	3.989 ms	7 - 130 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® X Elite	2.926 ms	25 - 25 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	2.164 ms	0 - 181 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® QCS6490	109.457 ms	55 - 117 MB	CPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	2.996 ms	0 - 32 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® QCS9075	3.375 ms	1 - 4 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® QCM6690	116.72 ms	60 - 70 MB	CPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	1.772 ms	0 - 131 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	117.799 ms	62 - 72 MB	CPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.513 ms	0 - 129 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® X Elite	15.375 ms	6 - 6 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.732 ms	6 - 83 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	51.115 ms	0 - 42 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	14.615 ms	6 - 8 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA8775P	19.101 ms	0 - 45 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® QCS9075	18.302 ms	6 - 14 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	26.626 ms	6 - 73 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA7255P	51.115 ms	0 - 42 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA8295P	20.99 ms	0 - 40 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	8.136 ms	6 - 52 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	6.624 ms	6 - 55 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® X Elite	5.018 ms	2 - 2 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.359 ms	2 - 103 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCS6490	15.019 ms	3 - 6 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	10.582 ms	2 - 56 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.677 ms	2 - 9 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA8775P	5.128 ms	2 - 57 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCS9075	6.849 ms	1 - 4 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCM6690	35.995 ms	2 - 189 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	7.19 ms	2 - 102 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA7255P	10.582 ms	2 - 56 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA8295P	6.557 ms	1 - 53 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.271 ms	2 - 52 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.278 ms	2 - 184 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.829 ms	2 - 56 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	9.767 ms	0 - 159 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® QCS8275 (Proxy)	51.077 ms	1 - 79 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® QCS8550 (Proxy)	14.781 ms	1 - 3 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA8775P	87.312 ms	1 - 79 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® QCS9075	17.611 ms	0 - 60 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® QCS8450 (Proxy)	26.15 ms	1 - 150 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA7255P	51.077 ms	1 - 79 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA8295P	21.057 ms	1 - 68 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	8.178 ms	1 - 84 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	6.661 ms	1 - 87 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	1.987 ms	0 - 104 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCS6490	9.502 ms	0 - 30 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	7.256 ms	0 - 52 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	2.779 ms	0 - 1 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA8775P	13.135 ms	0 - 53 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCS9075	3.252 ms	0 - 30 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCM6690	31.133 ms	0 - 188 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	3.552 ms	0 - 100 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA7255P	7.256 ms	0 - 52 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA8295P	4.414 ms	0 - 50 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	1.499 ms	0 - 53 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	4.107 ms	0 - 180 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.294 ms	0 - 53 MB	NPU

License

• The license for the original implementation of FFNet-78S-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.