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 2.45, 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

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:

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

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
DDRNet23-Slim	ONNX	float	Snapdragon® X2 Elite	11.729 ms	24 - 24 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® X Elite	27.603 ms	24 - 24 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Gen 3 Mobile	19.74 ms	29 - 298 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Gen 1 Mobile	45.616 ms	29 - 300 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	34.693 ms	24 - 51 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	28.482 ms	24 - 40 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® QCS8450	45.616 ms	29 - 300 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	36.24 ms	24 - 51 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	27.603 ms	24 - 24 MB	NPU
DDRNet23-Slim	ONNX	float	Qualcomm® Dragonwing™ Q-8750	15.035 ms	7 - 194 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Elite Mobile	15.035 ms	7 - 194 MB	NPU
DDRNet23-Slim	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	11.385 ms	5 - 219 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® X2 Elite	43.742 ms	7 - 7 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® X Elite	56.005 ms	7 - 7 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	39.531 ms	7 - 252 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	55.339 ms	8 - 252 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ QCS6490	153.412 ms	6 - 9 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-8275	50.089 ms	6 - 10 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	53.543 ms	6 - 15 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® QCS8450	55.339 ms	8 - 252 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-9075	52.475 ms	6 - 9 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-X7181	56.005 ms	7 - 7 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ Q-6690	196.27 ms	5 - 205 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ Q-7790	54.203 ms	7 - 208 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Qualcomm® Dragonwing™ Q-8750	40.977 ms	1 - 199 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Elite Mobile	40.977 ms	1 - 199 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	43.376 ms	1 - 204 MB	NPU
DDRNet23-Slim	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	54.203 ms	7 - 208 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® X2 Elite	11.756 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® X Elite	33.885 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	22.414 ms	24 - 298 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	66.757 ms	21 - 297 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	38.728 ms	24 - 53 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	98.414 ms	24 - 221 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	32.978 ms	24 - 26 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8775P	40.443 ms	24 - 222 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8650P	40.443 ms	24 - 222 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8255P	40.443 ms	24 - 222 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® QCS8450	66.757 ms	21 - 297 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	41.021 ms	24 - 52 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	33.885 ms	24 - 24 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	15.682 ms	24 - 243 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA7255P	98.414 ms	24 - 221 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Qualcomm® SA8295P	43.528 ms	24 - 231 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Elite Mobile	15.682 ms	24 - 243 MB	NPU
DDRNet23-Slim	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	10.532 ms	23 - 260 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® X2 Elite	29.342 ms	6 - 6 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® X Elite	59.018 ms	6 - 6 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	41.739 ms	6 - 253 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	80.424 ms	7 - 254 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-8275	54.923 ms	4 - 13 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-8275	107.928 ms	6 - 205 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	56.042 ms	6 - 113 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA8775P	56.911 ms	6 - 206 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA8650P	56.911 ms	6 - 206 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA8255P	56.911 ms	6 - 206 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® QCS8450	80.424 ms	7 - 254 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	57.728 ms	6 - 14 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	59.018 ms	6 - 6 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	39.556 ms	6 - 223 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA7255P	107.928 ms	6 - 205 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Qualcomm® SA8295P	66.574 ms	6 - 208 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	39.556 ms	6 - 223 MB	NPU
DDRNet23-Slim	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	28.748 ms	6 - 240 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	22.388 ms	1 - 281 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	66.52 ms	3 - 287 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	38.285 ms	0 - 41 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	98.304 ms	0 - 201 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	33.209 ms	2 - 5 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8775P	40.406 ms	3 - 205 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8650P	40.406 ms	3 - 205 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8255P	40.406 ms	3 - 205 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® QCS8450	66.52 ms	3 - 287 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	41.162 ms	0 - 41 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	15.599 ms	1 - 229 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA7255P	98.304 ms	0 - 201 MB	NPU
DDRNet23-Slim	TFLITE	float	Qualcomm® SA8295P	43.452 ms	2 - 214 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Elite Mobile	15.599 ms	1 - 229 MB	NPU
DDRNet23-Slim	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	10.444 ms	2 - 242 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