Modelpublic

qualcomm/FFNet-78S-LowRes

sourceHugging Facebsd-3-clauseupdated 10d agoView on Hugging Face

2likes2kdownloads

Model Card

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 2.45, ONNX Runtime 1.27.1	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit [FFNet-78S-LowRes on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/ffnet_78s_lowres).

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: Modelusecase.semantic_segmentation

Model Stats:

—Input resolution: 1024x512
—Model checkpoint: ffnet78SBCCcityscapesstatedictquartspre_down
—Model size (float): 102 MB
—Model size (w8a8): 26.0 MB
—Number of output classes: 19
—Number of parameters: 26.8M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FFNet-78S-LowRes	ONNX	float	Snapdragon® X2 Elite	35.709 ms	10 - 10 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® X Elite	20.101 ms	47 - 47 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Gen 3 Mobile	15.193 ms	2 - 251 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Gen 1 Mobile	35.842 ms	11 - 243 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	39.359 ms	6 - 15 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	20.029 ms	0 - 52 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® QCS8450	35.842 ms	11 - 243 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	35.717 ms	6 - 15 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	20.101 ms	47 - 47 MB	NPU
FFNet-78S-LowRes	ONNX	float	Qualcomm® Dragonwing™ Q-8750	13.372 ms	8 - 194 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Elite Mobile	13.372 ms	8 - 194 MB	NPU
FFNet-78S-LowRes	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	8.202 ms	8 - 196 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® X2 Elite	2.719 ms	4 - 4 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® X Elite	5.862 ms	25 - 25 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	3.969 ms	3 - 257 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	8.17 ms	3 - 256 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ QCS6490	28.757 ms	2 - 5 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-8275	5.94 ms	2 - 6 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	5.571 ms	0 - 28 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® QCS8450	8.17 ms	3 - 256 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-9075	5.781 ms	2 - 5 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-X7181	5.862 ms	25 - 25 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ Q-6690	105.84 ms	3 - 263 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ Q-7790	12.354 ms	3 - 215 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Qualcomm® Dragonwing™ Q-8750	3.047 ms	0 - 195 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Elite Mobile	3.047 ms	0 - 195 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	2.31 ms	0 - 205 MB	NPU
FFNet-78S-LowRes	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	12.354 ms	3 - 215 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® X2 Elite	36.31 ms	6 - 6 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® X Elite	22.73 ms	6 - 6 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	15.743 ms	6 - 246 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	42.173 ms	6 - 235 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	40.132 ms	6 - 17 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	21.647 ms	6 - 271 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA8775P	36.276 ms	1 - 183 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA8650P	36.276 ms	1 - 183 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA8255P	36.276 ms	1 - 183 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® QCS8450	42.173 ms	6 - 235 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	36.204 ms	8 - 18 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	22.73 ms	6 - 6 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	14.261 ms	6 - 196 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA7255P	131.844 ms	2 - 184 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Qualcomm® SA8295P	38.171 ms	0 - 177 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Elite Mobile	14.261 ms	6 - 196 MB	NPU
FFNet-78S-LowRes	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	8.57 ms	6 - 202 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® X2 Elite	3.442 ms	2 - 2 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® X Elite	7.657 ms	2 - 2 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	4.879 ms	0 - 128 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	8.581 ms	2 - 131 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS6490	31.069 ms	0 - 4 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-8275	7.402 ms	2 - 6 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	7.007 ms	2 - 4 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA8775P	7.375 ms	2 - 82 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA8650P	7.375 ms	2 - 82 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA8255P	7.375 ms	2 - 82 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® QCS8450	8.581 ms	2 - 131 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	7.115 ms	3 - 7 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	7.657 ms	2 - 2 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-6690	108.362 ms	2 - 205 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-7790	13.535 ms	2 - 216 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	3.537 ms	2 - 189 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA7255P	20.488 ms	2 - 80 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Qualcomm® SA8295P	11.568 ms	1 - 77 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	3.537 ms	2 - 189 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	2.687 ms	2 - 200 MB	NPU
FFNet-78S-LowRes	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	13.535 ms	2 - 216 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	15.704 ms	2 - 320 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	42.189 ms	3 - 310 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	40.017 ms	0 - 63 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	21.64 ms	2 - 5 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA8775P	36.349 ms	0 - 215 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA8650P	36.349 ms	0 - 215 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA8255P	36.349 ms	0 - 215 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® QCS8450	42.189 ms	3 - 310 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	36.233 ms	0 - 63 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	14.556 ms	2 - 226 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA7255P	131.865 ms	3 - 218 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Qualcomm® SA8295P	38.244 ms	3 - 214 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Elite Mobile	14.556 ms	2 - 226 MB	NPU
FFNet-78S-LowRes	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	8.581 ms	2 - 227 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	3.618 ms	1 - 131 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	6.744 ms	1 - 133 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS6490	26.769 ms	1 - 31 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-8275	5.766 ms	1 - 32 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.972 ms	1 - 3 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA8775P	5.497 ms	1 - 79 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA8650P	5.497 ms	1 - 79 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA8255P	5.497 ms	1 - 79 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® QCS8450	6.744 ms	1 - 133 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-9075	5.261 ms	1 - 31 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-6690	100.913 ms	1 - 201 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-7790	11.704 ms	1 - 211 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-8750	2.796 ms	0 - 187 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA7255P	17.357 ms	1 - 78 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Qualcomm® SA8295P	9.623 ms	1 - 75 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	2.796 ms	0 - 187 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	2.122 ms	0 - 195 MB	NPU
FFNet-78S-LowRes	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	11.704 ms	1 - 211 MB	NPU

License

—The license for the original implementation of FFNet-78S-LowRes can be found here.

References

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.