Hugging Face's logo

qualcomm
/
FFNet-40S

Image Segmentation
PyTorch
real_time
android
Model card Files
xet
 Community

FFNet-40S: Optimized for Qualcomm Devices

FFNet-40S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.

This is based on the implementation of FFNet-40S 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
TFLITE w8a8 Universal QAIRT 2.45 Download

For more device-specific assets and performance metrics, visit FFNet-40S 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 FFNet-40S on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

  • Input resolution: 2048x1024
  • Model checkpoint: ffnet40S_dBBB_cityscapes_state_dict_quarts
  • Model size (float): 53.1 MB
  • Model size (w8a8): 13.5 MB
  • Number of output classes: 19
  • Number of parameters: 13.9M

Performance Summary

Model Runtime Precision Chipset Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit
FFNet-40S ONNX float Snapdragon® X2 Elite 14.134 ms 24 - 24 MB NPU
FFNet-40S ONNX float Snapdragon® X Elite 31.242 ms 24 - 24 MB NPU
FFNet-40S ONNX float Snapdragon® 8 Gen 3 Mobile 22.444 ms 30 - 316 MB NPU
FFNet-40S ONNX float Snapdragon® 8 Gen 1 Mobile 52.058 ms 29 - 289 MB NPU
FFNet-40S ONNX float Qualcomm® Dragonwing™ IQ-8275 43.949 ms 24 - 28 MB NPU
FFNet-40S ONNX float Qualcomm® Dragonwing™ QCS8550 (Proxy) 31.836 ms 24 - 55 MB NPU
FFNet-40S ONNX float Qualcomm® QCS8450 52.058 ms 29 - 289 MB NPU
FFNet-40S ONNX float Qualcomm® Dragonwing™ IQ-9075 44.366 ms 24 - 27 MB NPU
FFNet-40S ONNX float Qualcomm® Dragonwing™ IQ-X7181 31.242 ms 24 - 24 MB NPU
FFNet-40S ONNX float Qualcomm® Dragonwing™ Q-8750 17.596 ms 5 - 191 MB NPU
FFNet-40S ONNX float Snapdragon® 8 Elite Mobile 17.596 ms 5 - 191 MB NPU
FFNet-40S ONNX float Snapdragon® 8 Elite Gen 5 Mobile 13.49 ms 5 - 216 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® X2 Elite 6.923 ms 7 - 7 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® X Elite 11.702 ms 9 - 9 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® 8 Gen 3 Mobile 7.38 ms 6 - 251 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® 8 Gen 1 Mobile 15.808 ms 7 - 254 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ QCS6490 48.677 ms 6 - 9 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ IQ-8275 10.821 ms 6 - 10 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ QCS8550 (Proxy) 11.236 ms 0 - 11 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® QCS8450 15.808 ms 7 - 254 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ IQ-9075 12.074 ms 6 - 9 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ IQ-X7181 11.702 ms 9 - 9 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ Q-6690 112.391 ms 5 - 242 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ Q-7790 14.92 ms 7 - 224 MB NPU
FFNet-40S ONNX w8a8 Qualcomm® Dragonwing™ Q-8750 7.664 ms 2 - 192 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® 8 Elite Mobile 7.664 ms 2 - 192 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® 8 Elite Gen 5 Mobile 6.618 ms 1 - 196 MB NPU
FFNet-40S ONNX w8a8 Snapdragon® 7 Gen 4 Mobile 14.92 ms 7 - 224 MB NPU
FFNet-40S QNN_DLC float Snapdragon® X2 Elite 14.525 ms 24 - 24 MB NPU
FFNet-40S QNN_DLC float Snapdragon® X Elite 37.758 ms 24 - 24 MB NPU
FFNet-40S QNN_DLC float Snapdragon® 8 Gen 3 Mobile 24.951 ms 23 - 290 MB NPU
FFNet-40S QNN_DLC float Snapdragon® 8 Gen 1 Mobile 63.55 ms 23 - 289 MB NPU
FFNet-40S QNN_DLC float Qualcomm® Dragonwing™ IQ-8275 48.511 ms 24 - 53 MB NPU
FFNet-40S QNN_DLC float Qualcomm® Dragonwing™ QCS8550 (Proxy) 36.722 ms 24 - 26 MB NPU
FFNet-40S QNN_DLC float Qualcomm® SA8775P 48.929 ms 24 - 219 MB NPU
FFNet-40S QNN_DLC float Qualcomm® SA8650P 48.929 ms 24 - 219 MB NPU
FFNet-40S QNN_DLC float Qualcomm® SA8255P 48.929 ms 24 - 219 MB NPU
FFNet-40S QNN_DLC float Qualcomm® QCS8450 63.55 ms 23 - 289 MB NPU
FFNet-40S QNN_DLC float Qualcomm® Dragonwing™ IQ-9075 50.027 ms 24 - 52 MB NPU
FFNet-40S QNN_DLC float Qualcomm® Dragonwing™ IQ-X7181 37.758 ms 24 - 24 MB NPU
FFNet-40S QNN_DLC float Qualcomm® Dragonwing™ Q-8750 18.104 ms 24 - 247 MB NPU
FFNet-40S QNN_DLC float Qualcomm® SA7255P 135.506 ms 24 - 219 MB NPU
FFNet-40S QNN_DLC float Qualcomm® SA8295P 53.665 ms 24 - 224 MB NPU
FFNet-40S QNN_DLC float Snapdragon® 8 Elite Mobile 18.104 ms 24 - 247 MB NPU
FFNet-40S QNN_DLC float Snapdragon® 8 Elite Gen 5 Mobile 12.712 ms 18 - 258 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® X2 Elite 6.066 ms 6 - 6 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® X Elite 15.944 ms 6 - 6 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® 8 Gen 3 Mobile 10.515 ms 6 - 252 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® 8 Gen 1 Mobile 21.381 ms 6 - 252 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ QCS6490 55.597 ms 6 - 14 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ IQ-8275 14.828 ms 6 - 15 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ QCS8550 (Proxy) 15.052 ms 6 - 8 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® SA8775P 15.675 ms 6 - 203 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® SA8650P 15.675 ms 6 - 203 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® SA8255P 15.675 ms 6 - 203 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® QCS8450 21.381 ms 6 - 252 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ IQ-9075 16.095 ms 6 - 14 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ IQ-X7181 15.944 ms 6 - 6 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ Q-6690 128.297 ms 6 - 238 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ Q-7790 18.895 ms 6 - 225 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® Dragonwing™ Q-8750 7.068 ms 3 - 214 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® SA7255P 32.92 ms 6 - 202 MB NPU
FFNet-40S QNN_DLC w8a8 Qualcomm® SA8295P 20.083 ms 6 - 204 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® 8 Elite Mobile 7.068 ms 3 - 214 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® 8 Elite Gen 5 Mobile 5.286 ms 6 - 232 MB NPU
FFNet-40S QNN_DLC w8a8 Snapdragon® 7 Gen 4 Mobile 18.895 ms 6 - 225 MB NPU
FFNet-40S TFLITE float Snapdragon® 8 Gen 3 Mobile 25.028 ms 1 - 304 MB NPU
FFNet-40S TFLITE float Snapdragon® 8 Gen 1 Mobile 63.411 ms 2 - 304 MB NPU
FFNet-40S TFLITE float Qualcomm® Dragonwing™ IQ-8275 48.572 ms 0 - 57 MB NPU
FFNet-40S TFLITE float Qualcomm® Dragonwing™ QCS8550 (Proxy) 36.278 ms 3 - 5 MB NPU
FFNet-40S TFLITE float Qualcomm® SA8775P 49.002 ms 2 - 211 MB NPU
FFNet-40S TFLITE float Qualcomm® SA8650P 49.002 ms 2 - 211 MB NPU
FFNet-40S TFLITE float Qualcomm® SA8255P 49.002 ms 2 - 211 MB NPU
FFNet-40S TFLITE float Qualcomm® QCS8450 63.411 ms 2 - 304 MB NPU
FFNet-40S TFLITE float Qualcomm® Dragonwing™ IQ-9075 49.94 ms 0 - 56 MB NPU
FFNet-40S TFLITE float Qualcomm® Dragonwing™ Q-8750 18.119 ms 2 - 240 MB NPU
FFNet-40S TFLITE float Qualcomm® SA7255P 135.531 ms 3 - 211 MB NPU
FFNet-40S TFLITE float Qualcomm® SA8295P 53.667 ms 0 - 215 MB NPU
FFNet-40S TFLITE float Snapdragon® 8 Elite Mobile 18.119 ms 2 - 240 MB NPU
FFNet-40S TFLITE float Snapdragon® 8 Elite Gen 5 Mobile 12.812 ms 2 - 253 MB NPU
FFNet-40S TFLITE w8a8 Snapdragon® 8 Gen 3 Mobile 5.456 ms 0 - 247 MB NPU
FFNet-40S TFLITE w8a8 Snapdragon® 8 Gen 1 Mobile 12.361 ms 1 - 248 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ QCS6490 41.127 ms 1 - 23 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ IQ-8275 8.033 ms 1 - 24 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ QCS8550 (Proxy) 7.495 ms 1 - 20 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® SA8775P 8.267 ms 1 - 197 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® SA8650P 8.267 ms 1 - 197 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® SA8255P 8.267 ms 1 - 197 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® QCS8450 12.361 ms 1 - 248 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ IQ-9075 8.032 ms 1 - 23 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ Q-6690 102.581 ms 1 - 231 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ Q-7790 11.536 ms 1 - 215 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® Dragonwing™ Q-8750 3.995 ms 1 - 212 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® SA7255P 20.068 ms 1 - 196 MB NPU
FFNet-40S TFLITE w8a8 Qualcomm® SA8295P 11.694 ms 1 - 201 MB NPU
FFNet-40S TFLITE w8a8 Snapdragon® 8 Elite Mobile 3.995 ms 1 - 212 MB NPU
FFNet-40S TFLITE w8a8 Snapdragon® 8 Elite Gen 5 Mobile 2.977 ms 1 - 226 MB NPU
FFNet-40S TFLITE w8a8 Snapdragon® 7 Gen 4 Mobile 11.536 ms 1 - 215 MB NPU

License

  • The license for the original implementation of FFNet-40S can be found here.

References

Community

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Image Segmentation
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for qualcomm/FFNet-40S