FCN-ResNet50: Optimized for Qualcomm Devices

FCN_ResNet50 is a machine learning model that can segment images from the COCO dataset. It uses ResNet50 as a backbone.

This is based on the implementation of FCN-ResNet50 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 1.25.0	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.25.0	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 FCN-ResNet50 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 FCN-ResNet50 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: COCO_WITH_VOC_LABELS_V1
• Input resolution: 224x224
• Number of output classes: 21
• Number of parameters: 33.0M
• Model size (float): 126 MB
• Model size (w8a8): 32.2 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
FCN-ResNet50	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	22.985 ms	4 - 321 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® X2 Elite	23.342 ms	209 - 209 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® X Elite	43.052 ms	146 - 146 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	32.884 ms	3 - 386 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS8550 (Proxy)	44.034 ms	0 - 90 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	26.527 ms	1 - 304 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS9075	75.206 ms	3 - 48 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS8750	26.527 ms	1 - 304 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS7181	43.052 ms	146 - 146 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.059 ms	1 - 249 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® X2 Elite	7.123 ms	211 - 211 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® X Elite	13.681 ms	180 - 180 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	10.005 ms	1 - 282 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS6490	907.536 ms	69 - 111 MB	CPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	13.569 ms	0 - 56 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	705.289 ms	48 - 57 MB	CPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCM6690	850.814 ms	42 - 51 MB	CPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.576 ms	1 - 204 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS9075	15.136 ms	1 - 46 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS7790	705.289 ms	48 - 57 MB	CPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS8750	8.576 ms	1 - 204 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS7181	13.681 ms	180 - 180 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	21.569 ms	3 - 335 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® X2 Elite	22.936 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® X Elite	44.74 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	33.254 ms	0 - 376 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8275	273.663 ms	2 - 310 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	44.393 ms	3 - 26 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8775P	72.869 ms	1 - 309 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8650P	72.869 ms	1 - 309 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8255P	72.869 ms	1 - 309 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	84.845 ms	1 - 269 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8295P	77.81 ms	0 - 218 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	26.639 ms	0 - 318 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA7255P	273.663 ms	2 - 310 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS9075	78.335 ms	3 - 8 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8750	26.639 ms	0 - 318 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS7181	44.74 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.43 ms	1 - 242 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® X2 Elite	7.891 ms	1 - 1 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® X Elite	15.184 ms	1 - 1 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.235 ms	1 - 263 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS6490	93.671 ms	1 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8275	39.51 ms	1 - 207 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	14.923 ms	1 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8775P	14.696 ms	1 - 216 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8650P	14.696 ms	1 - 216 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8255P	14.696 ms	1 - 216 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA7255P	39.51 ms	1 - 207 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	25.943 ms	1 - 280 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8295P	21.926 ms	1 - 210 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCM6690	382.682 ms	1 - 384 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.926 ms	1 - 202 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS9075	17.496 ms	1 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	24.528 ms	1 - 265 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS7790	25.943 ms	1 - 280 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8750	8.926 ms	1 - 202 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS7181	15.184 ms	1 - 1 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	21.104 ms	0 - 368 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	33.216 ms	0 - 428 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8275	273.711 ms	1 - 353 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8550 (Proxy)	44.228 ms	0 - 3 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8775P	72.824 ms	1 - 354 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8650P	72.824 ms	1 - 354 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8255P	72.824 ms	1 - 354 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8450 (Proxy)	84.178 ms	1 - 321 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8295P	77.788 ms	0 - 268 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	26.819 ms	0 - 357 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA7255P	273.711 ms	1 - 353 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS9075	78.275 ms	0 - 71 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8750	26.819 ms	0 - 357 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.35 ms	0 - 243 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	10.924 ms	0 - 266 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS6490	95.129 ms	0 - 39 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8275	38.283 ms	0 - 206 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	14.34 ms	0 - 2 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8775P	14.701 ms	5 - 212 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8650P	14.701 ms	5 - 212 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8255P	14.701 ms	5 - 212 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA7255P	38.283 ms	0 - 206 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	27.953 ms	0 - 280 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8295P	21.315 ms	0 - 210 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCM6690	438.041 ms	0 - 388 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.619 ms	0 - 198 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS9075	15.315 ms	0 - 35 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	23.299 ms	0 - 268 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS7790	27.953 ms	0 - 280 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8750	8.619 ms	0 - 198 MB	NPU

License

• The license for the original implementation of FCN-ResNet50 can be found here.

References

• Fully Convolutional Networks 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.