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 1.24.3	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 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	22.429 ms	4 - 326 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® X2 Elite	23.176 ms	63 - 63 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® X Elite	43.168 ms	62 - 62 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	32.577 ms	0 - 382 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS8550 (Proxy)	43.03 ms	3 - 6 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS9075	75.156 ms	3 - 9 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	26.466 ms	1 - 297 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.055 ms	0 - 248 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® X2 Elite	7.077 ms	33 - 33 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® X Elite	13.821 ms	32 - 32 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	10.174 ms	0 - 279 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS6490	907.251 ms	68 - 111 MB	CPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	13.515 ms	1 - 259 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS9075	15.155 ms	1 - 4 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCM6690	843.043 ms	54 - 63 MB	CPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.58 ms	1 - 206 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	707.962 ms	80 - 88 MB	CPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	21.662 ms	3 - 331 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® X2 Elite	22.829 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® X Elite	44.195 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	33.24 ms	3 - 385 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	272.365 ms	1 - 304 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	44.069 ms	3 - 6 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8775P	72.498 ms	1 - 304 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS9075	78.663 ms	3 - 8 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	83.904 ms	1 - 275 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA7255P	272.365 ms	1 - 304 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8295P	77.755 ms	0 - 217 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	26.24 ms	3 - 322 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.542 ms	1 - 246 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® X2 Elite	7.864 ms	1 - 1 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® X Elite	15.175 ms	1 - 1 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.375 ms	1 - 267 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS6490	92.908 ms	1 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	39.543 ms	1 - 205 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	14.947 ms	1 - 197 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8775P	14.759 ms	1 - 215 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS9075	17.017 ms	3 - 5 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCM6690	392.044 ms	1 - 347 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	23.99 ms	1 - 267 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA7255P	39.543 ms	1 - 205 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8295P	22.112 ms	1 - 209 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.869 ms	1 - 202 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	26.065 ms	1 - 281 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	20.489 ms	0 - 355 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	33.281 ms	0 - 432 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8275 (Proxy)	272.269 ms	0 - 329 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8550 (Proxy)	44.565 ms	0 - 3 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8775P	72.513 ms	0 - 330 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS9075	78.557 ms	0 - 71 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8450 (Proxy)	83.942 ms	0 - 316 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA7255P	272.269 ms	0 - 329 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8295P	77.812 ms	0 - 246 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	27.329 ms	0 - 349 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.145 ms	0 - 247 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	11.09 ms	0 - 270 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS6490	94.653 ms	0 - 39 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	38.256 ms	0 - 206 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	14.359 ms	0 - 5 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8775P	14.754 ms	0 - 205 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS9075	14.952 ms	0 - 35 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCM6690	433.127 ms	0 - 352 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	23.27 ms	0 - 268 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA7255P	38.256 ms	0 - 206 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8295P	21.338 ms	0 - 209 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.553 ms	0 - 199 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	26.346 ms	0 - 281 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.