DDRNet23-Slim: Optimized for Qualcomm Devices

DDRNet23Slim is a machine learning model that segments an image into semantic classes, specifically designed for road-based scenes. It is designed for the application of self-driving cars.

This is based on the implementation of DDRNet23-Slim 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 DDRNet23-Slim 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 DDRNet23-Slim on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

Model checkpoint: DDRNet23s_imagenet.pth
Inference latency: RealTime
Input resolution: 2048x1024
Number of output classes: 19
Number of parameters: 6.13M
Model size (float): 21.7 MB
Model size (w8a8): 6.11 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
DDRNet23-Slim	ONNX	float	Snapdragon® X2 Elite	10.937 ms	22 - 22 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® X Elite	27.931 ms	24 - 24 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Gen 3 Mobile	19.68 ms	30 - 309 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS8550 (Proxy)	28.907 ms	24 - 27 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS9075	39.526 ms	24 - 51 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	13.576 ms	3 - 197 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	10.474 ms	30 - 259 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® X2 Elite	44.614 ms	109 - 109 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® X Elite	87.138 ms	109 - 109 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	43.588 ms	91 - 333 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS6490	299.103 ms	198 - 215 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	57.741 ms	86 - 90 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS9075	64.439 ms	87 - 89 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCM6690	266.03 ms	132 - 141 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	42.186 ms	81 - 269 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	250.763 ms	139 - 148 MB	CPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	43.559 ms	77 - 269 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® X2 Elite	11.824 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® X Elite	33.583 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	22.45 ms	22 - 307 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	98.319 ms	24 - 222 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	32.53 ms	24 - 224 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8775P	40.267 ms	24 - 223 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS9075	53.298 ms	24 - 52 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	67.424 ms	0 - 285 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA7255P	98.319 ms	24 - 222 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8295P	42.961 ms	24 - 229 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	15.218 ms	23 - 247 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	10.374 ms	12 - 248 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® X2 Elite	47.73 ms	6 - 6 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® X Elite	58.745 ms	6 - 6 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	42.01 ms	6 - 255 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	107.806 ms	6 - 204 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	56.2 ms	6 - 8 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA8775P	56.982 ms	6 - 205 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® QCS9075	60.328 ms	6 - 14 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	61.369 ms	6 - 256 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA7255P	107.806 ms	6 - 204 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA8295P	64.388 ms	6 - 208 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	40.548 ms	6 - 222 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	47.033 ms	6 - 238 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	22.502 ms	2 - 297 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8275 (Proxy)	98.293 ms	0 - 203 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8550 (Proxy)	32.944 ms	2 - 4 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8775P	40.147 ms	2 - 206 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS9075	53.893 ms	0 - 41 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8450 (Proxy)	67.224 ms	2 - 299 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA7255P	98.293 ms	0 - 203 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8295P	43.035 ms	2 - 216 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	15.22 ms	1 - 226 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	10.328 ms	2 - 242 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	36.953 ms	1 - 249 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® QCS6490	198.376 ms	9 - 77 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	95.138 ms	1 - 198 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	49.049 ms	1 - 4 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® SA8775P	49.608 ms	1 - 199 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® QCS9075	51.749 ms	1 - 15 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® QCM6690	223.252 ms	10 - 196 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	56.777 ms	0 - 251 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® SA7255P	95.138 ms	1 - 198 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Qualcomm® SA8295P	56.049 ms	1 - 203 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	65.119 ms	1 - 215 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	64.885 ms	9 - 209 MB	NPU
DDRNet23-Slim	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	44.865 ms	1 - 231 MB	NPU

License

The license for the original implementation of DDRNet23-Slim 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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Paper for qualcomm/DDRNet23-Slim

Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes

Paper • 2101.06085 • Published Jan 15, 2021