SalsaNext: Optimized for Qualcomm Devices
SalsaNext is a LiDAR-based model designed for efficient and accurate semantic segmentation.
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 1.25.0 | Download |
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | Download |
| QNN_DLC | float | Universal | QAIRT 2.45 | Download |
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | Download |
| TFLITE | float | Universal | QAIRT 2.45 | Download |
For more device-specific assets and performance metrics, visit SalsaNext 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 SalsaNext on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Model checkpoint: SalsaNext
- Input resolution: 1x5x64x2048
- Number of parameters: 6.71M
- Model size (float): 25.7 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| SalsaNext | ONNX | float | Snapdragon® X2 Elite | 13.787 ms | 210 - 210 MB | NPU |
| SalsaNext | ONNX | float | Snapdragon® X Elite | 32.372 ms | 147 - 147 MB | NPU |
| SalsaNext | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 25.895 ms | 27 - 350 MB | NPU |
| SalsaNext | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 66.246 ms | 24 - 362 MB | NPU |
| SalsaNext | ONNX | float | Qualcomm® QCS8550 (Proxy) | 34.124 ms | 13 - 42 MB | NPU |
| SalsaNext | ONNX | float | Qualcomm® QCS8450 | 66.246 ms | 24 - 362 MB | NPU |
| SalsaNext | ONNX | float | Snapdragon® 8 Elite Mobile | 19.076 ms | 23 - 294 MB | NPU |
| SalsaNext | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.533 ms | 23 - 313 MB | NPU |
| SalsaNext | ONNX | float | Qualcomm® QCS9075 | 37.818 ms | 23 - 68 MB | NPU |
| SalsaNext | ONNX | float | Qualcomm® QCS8750 | 19.076 ms | 23 - 294 MB | NPU |
| SalsaNext | ONNX | float | Qualcomm® QCS7181 | 32.372 ms | 147 - 147 MB | NPU |
| SalsaNext | ONNX | w8a16 | Snapdragon® X2 Elite | 16.632 ms | 212 - 212 MB | NPU |
| SalsaNext | ONNX | w8a16 | Snapdragon® X Elite | 33.555 ms | 148 - 148 MB | NPU |
| SalsaNext | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 27.069 ms | 15 - 264 MB | NPU |
| SalsaNext | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 35.279 ms | 10 - 25 MB | NPU |
| SalsaNext | ONNX | w8a16 | Qualcomm® QCS9075 | 34.524 ms | 15 - 60 MB | NPU |
| SalsaNext | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 18.451 ms | 13 - 201 MB | NPU |
| SalsaNext | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 23.522 ms | 13 - 206 MB | NPU |
| SalsaNext | ONNX | w8a16 | Qualcomm® QCS8750 | 23.522 ms | 13 - 206 MB | NPU |
| SalsaNext | ONNX | w8a16 | Qualcomm® QCS7181 | 33.555 ms | 148 - 148 MB | NPU |
| SalsaNext | QNN_DLC | float | Snapdragon® X2 Elite | 13.828 ms | 3 - 3 MB | NPU |
| SalsaNext | QNN_DLC | float | Snapdragon® X Elite | 31.896 ms | 3 - 3 MB | NPU |
| SalsaNext | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 25.508 ms | 0 - 330 MB | NPU |
| SalsaNext | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 65.688 ms | 2 - 349 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® QCS8275 | 118.658 ms | 1 - 247 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 34.488 ms | 3 - 5 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® QCS8450 | 65.688 ms | 2 - 349 MB | NPU |
| SalsaNext | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 18.523 ms | 0 - 268 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® SA8295P | 51.611 ms | 0 - 259 MB | NPU |
| SalsaNext | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.198 ms | 3 - 294 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® SA7255P | 118.658 ms | 1 - 247 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® QCS9075 | 37.168 ms | 3 - 17 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® QCS8750 | 18.523 ms | 0 - 268 MB | NPU |
| SalsaNext | QNN_DLC | float | Qualcomm® QCS7181 | 31.896 ms | 3 - 3 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 18.51 ms | 1 - 1 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Snapdragon® X Elite | 40.701 ms | 1 - 1 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 33.576 ms | 0 - 366 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 71.509 ms | 1 - 420 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® QCS8275 | 76.363 ms | 1 - 277 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 40.697 ms | 1 - 283 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 71.509 ms | 1 - 420 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 47.8 ms | 1 - 9 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 19.639 ms | 1 - 349 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 27.631 ms | 1 - 294 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® SA8295P | 59.157 ms | 1 - 328 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® SA7255P | 76.363 ms | 1 - 277 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® QCS8750 | 27.631 ms | 1 - 294 MB | NPU |
| SalsaNext | QNN_DLC | w8a16 | Qualcomm® QCS7181 | 40.701 ms | 1 - 1 MB | NPU |
| SalsaNext | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 25.597 ms | 0 - 336 MB | NPU |
| SalsaNext | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 65.508 ms | 10 - 360 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® QCS8275 | 118.628 ms | 10 - 257 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 33.684 ms | 10 - 225 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® SA8775P | 230.405 ms | 10 - 40 MB | GPU |
| SalsaNext | TFLITE | float | Qualcomm® SA8650P | 230.405 ms | 10 - 40 MB | GPU |
| SalsaNext | TFLITE | float | Qualcomm® SA8255P | 230.405 ms | 10 - 40 MB | GPU |
| SalsaNext | TFLITE | float | Qualcomm® QCS8450 | 65.508 ms | 10 - 360 MB | NPU |
| SalsaNext | TFLITE | float | Snapdragon® 8 Elite Mobile | 18.549 ms | 9 - 272 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® SA8295P | 51.663 ms | 10 - 272 MB | NPU |
| SalsaNext | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.199 ms | 10 - 299 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® SA7255P | 118.628 ms | 10 - 257 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® QCS9075 | 36.911 ms | 10 - 39 MB | NPU |
| SalsaNext | TFLITE | float | Qualcomm® QCS8750 | 18.549 ms | 9 - 272 MB | NPU |
License
- The license for the original implementation of SalsaNext can be found here.
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.
