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 1.24.1	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.17.0	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® X Elite	43.191 ms	62 - 62 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	33.044 ms	0 - 386 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS8550 (Proxy)	43.554 ms	3 - 383 MB	NPU
FCN-ResNet50	ONNX	float	Qualcomm® QCS9075	75.301 ms	3 - 9 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	26.715 ms	2 - 309 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	22.44 ms	4 - 325 MB	NPU
FCN-ResNet50	ONNX	float	Snapdragon® X2 Elite	23.214 ms	63 - 63 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® X Elite	13.847 ms	32 - 32 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	10.161 ms	1 - 280 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS6490	915.953 ms	69 - 113 MB	CPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	13.499 ms	1 - 4 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCS9075	15.126 ms	1 - 4 MB	NPU
FCN-ResNet50	ONNX	w8a8	Qualcomm® QCM6690	844.226 ms	63 - 72 MB	CPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.557 ms	1 - 203 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	711.735 ms	66 - 75 MB	CPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.058 ms	1 - 248 MB	NPU
FCN-ResNet50	ONNX	w8a8	Snapdragon® X2 Elite	7.098 ms	33 - 33 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® X Elite	44.233 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	33.588 ms	3 - 387 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	272.33 ms	1 - 304 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	43.876 ms	3 - 5 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8775P	72.461 ms	1 - 304 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS9075	78.672 ms	3 - 8 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	83.781 ms	0 - 273 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA7255P	272.33 ms	1 - 304 MB	NPU
FCN-ResNet50	QNN_DLC	float	Qualcomm® SA8295P	77.696 ms	0 - 218 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	26.543 ms	0 - 318 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	21.396 ms	3 - 331 MB	NPU
FCN-ResNet50	QNN_DLC	float	Snapdragon® X2 Elite	22.797 ms	3 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® X Elite	15.188 ms	1 - 1 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	11.457 ms	0 - 265 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS6490	92.768 ms	1 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	39.523 ms	1 - 205 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	14.945 ms	0 - 16 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8775P	64.67 ms	1 - 215 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS9075	16.881 ms	1 - 3 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCM6690	405.918 ms	1 - 347 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	23.823 ms	1 - 266 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA7255P	39.523 ms	1 - 205 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Qualcomm® SA8295P	22.143 ms	1 - 209 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.873 ms	1 - 200 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	26.146 ms	1 - 281 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.558 ms	1 - 245 MB	NPU
FCN-ResNet50	QNN_DLC	w8a8	Snapdragon® X2 Elite	7.826 ms	1 - 1 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	33.402 ms	0 - 431 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8275 (Proxy)	272.27 ms	0 - 332 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8550 (Proxy)	44.378 ms	0 - 3 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8775P	72.555 ms	0 - 333 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS9075	77.94 ms	0 - 71 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® QCS8450 (Proxy)	83.943 ms	0 - 318 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA7255P	272.27 ms	0 - 332 MB	NPU
FCN-ResNet50	TFLITE	float	Qualcomm® SA8295P	77.823 ms	0 - 246 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	27.098 ms	0 - 343 MB	NPU
FCN-ResNet50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	20.491 ms	0 - 355 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	11.181 ms	0 - 272 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS6490	94.818 ms	0 - 39 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	38.251 ms	0 - 205 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	14.418 ms	0 - 2 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8775P	14.773 ms	0 - 205 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS9075	15.44 ms	0 - 35 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCM6690	437.87 ms	0 - 352 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	23.589 ms	0 - 268 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA7255P	38.251 ms	0 - 205 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Qualcomm® SA8295P	21.346 ms	0 - 209 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	8.546 ms	0 - 200 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	26.736 ms	0 - 282 MB	NPU
FCN-ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	7.138 ms	0 - 247 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.