---
library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: image-segmentation

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/web-assets/model_demo.png)

# 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](https://github.com/Qualcomm-AI-research/FFNet).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/ffnet_40s) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) 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](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/releases/v0.59.0/ffnet_40s-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/releases/v0.59.0/ffnet_40s-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/releases/v0.59.0/ffnet_40s-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/releases/v0.59.0/ffnet_40s-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/releases/v0.59.0/ffnet_40s-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_40s/releases/v0.59.0/ffnet_40s-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[FFNet-40S on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/ffnet_40s)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/ffnet_40s) 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](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/ffnet_40s) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

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

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| FFNet-40S | ONNX | float | Snapdragon® X2 Elite | 14.06 ms | 24 - 24 MB | NPU
| FFNet-40S | ONNX | float | Snapdragon® X Elite | 31.528 ms | 24 - 24 MB | NPU
| FFNet-40S | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 22.232 ms | 30 - 316 MB | NPU
| FFNet-40S | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 52.29 ms | 29 - 288 MB | NPU
| FFNet-40S | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 32.232 ms | 24 - 67 MB | NPU
| FFNet-40S | ONNX | float | Qualcomm® QCS8450 | 52.29 ms | 29 - 288 MB | NPU
| FFNet-40S | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 44.419 ms | 24 - 27 MB | NPU
| FFNet-40S | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 31.528 ms | 24 - 24 MB | NPU
| FFNet-40S | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 17.695 ms | 7 - 192 MB | NPU
| FFNet-40S | ONNX | float | Snapdragon® 8 Elite Mobile | 17.695 ms | 7 - 192 MB | NPU
| FFNet-40S | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.504 ms | 5 - 217 MB | NPU
| FFNet-40S | ONNX | w8a8 | Snapdragon® X2 Elite | 6.933 ms | 7 - 7 MB | NPU
| FFNet-40S | ONNX | w8a8 | Snapdragon® X Elite | 11.776 ms | 9 - 9 MB | NPU
| FFNet-40S | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 7.39 ms | 7 - 253 MB | NPU
| FFNet-40S | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 15.759 ms | 7 - 254 MB | NPU
| FFNet-40S | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 48.787 ms | 6 - 9 MB | NPU
| FFNet-40S | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 11.251 ms | 0 - 12 MB | NPU
| FFNet-40S | ONNX | w8a8 | Qualcomm® QCS8450 | 15.759 ms | 7 - 254 MB | NPU
| FFNet-40S | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 11.778 ms | 6 - 9 MB | NPU
| FFNet-40S | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 11.776 ms | 9 - 9 MB | NPU
| FFNet-40S | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 7.67 ms | 1 - 191 MB | NPU
| FFNet-40S | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 7.67 ms | 1 - 191 MB | NPU
| FFNet-40S | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 6.634 ms | 2 - 195 MB | NPU
| FFNet-40S | QNN_DLC | float | Snapdragon® X2 Elite | 14.618 ms | 24 - 24 MB | NPU
| FFNet-40S | QNN_DLC | float | Snapdragon® X Elite | 37.705 ms | 24 - 24 MB | NPU
| FFNet-40S | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 24.955 ms | 24 - 293 MB | NPU
| FFNet-40S | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 63.448 ms | 24 - 290 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 135.537 ms | 24 - 219 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 36.137 ms | 19 - 242 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® SA8775P | 48.907 ms | 24 - 219 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® SA8650P | 48.907 ms | 24 - 219 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® SA8255P | 48.907 ms | 24 - 219 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® QCS8450 | 63.448 ms | 24 - 290 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 49.753 ms | 24 - 52 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 37.705 ms | 24 - 24 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 18.092 ms | 0 - 224 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® SA7255P | 135.537 ms | 24 - 219 MB | NPU
| FFNet-40S | QNN_DLC | float | Qualcomm® SA8295P | 53.688 ms | 24 - 224 MB | NPU
| FFNet-40S | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 18.092 ms | 0 - 224 MB | NPU
| FFNet-40S | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 12.583 ms | 20 - 261 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 6.104 ms | 6 - 6 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® X Elite | 16.175 ms | 6 - 6 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 10.502 ms | 6 - 252 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 21.31 ms | 6 - 252 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 55.759 ms | 7 - 15 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 32.935 ms | 6 - 202 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.051 ms | 6 - 8 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® SA8775P | 15.703 ms | 6 - 203 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® SA8650P | 15.703 ms | 6 - 203 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® SA8255P | 15.703 ms | 6 - 203 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 21.31 ms | 6 - 252 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 15.635 ms | 6 - 14 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 16.175 ms | 6 - 6 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 131.656 ms | 6 - 239 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 19.085 ms | 6 - 223 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 7.074 ms | 6 - 216 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® SA7255P | 32.935 ms | 6 - 202 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Qualcomm® SA8295P | 20.082 ms | 6 - 204 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 7.074 ms | 6 - 216 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 5.301 ms | 6 - 233 MB | NPU
| FFNet-40S | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 19.085 ms | 6 - 223 MB | NPU
| FFNet-40S | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 25.019 ms | 1 - 305 MB | NPU
| FFNet-40S | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 63.785 ms | 0 - 300 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 135.541 ms | 2 - 210 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 36.242 ms | 2 - 5 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® SA8775P | 49.038 ms | 3 - 212 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® SA8650P | 49.038 ms | 3 - 212 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® SA8255P | 49.038 ms | 3 - 212 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® QCS8450 | 63.785 ms | 0 - 300 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 50.268 ms | 0 - 56 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 18.056 ms | 0 - 236 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® SA7255P | 135.541 ms | 2 - 210 MB | NPU
| FFNet-40S | TFLITE | float | Qualcomm® SA8295P | 53.682 ms | 3 - 218 MB | NPU
| FFNet-40S | TFLITE | float | Snapdragon® 8 Elite Mobile | 18.056 ms | 0 - 236 MB | NPU
| FFNet-40S | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 12.69 ms | 2 - 254 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 5.46 ms | 1 - 248 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 12.396 ms | 1 - 247 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 40.735 ms | 1 - 23 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 20.07 ms | 1 - 196 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.456 ms | 1 - 4 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® SA8775P | 8.263 ms | 1 - 197 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® SA8650P | 8.263 ms | 1 - 197 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® SA8255P | 8.263 ms | 1 - 197 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® QCS8450 | 12.396 ms | 1 - 247 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 8.027 ms | 1 - 23 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 102.048 ms | 0 - 230 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 11.483 ms | 1 - 216 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 3.985 ms | 0 - 208 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® SA7255P | 20.07 ms | 1 - 196 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Qualcomm® SA8295P | 11.663 ms | 1 - 200 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 3.985 ms | 0 - 208 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 2.996 ms | 1 - 225 MB | NPU
| FFNet-40S | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 11.483 ms | 1 - 216 MB | NPU

## License
* The license for the original implementation of FFNet-40S can be found
  [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).

## References
* [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236)
* [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).