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

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_54s/web-assets/model_demo.png)

# FFNet-54S: Optimized for Qualcomm Devices

FFNet-54S 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-54S 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/main/src/qai_hub_models/models/ffnet_54s) 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 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_54s/releases/v0.51.0/ffnet_54s-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_54s/releases/v0.51.0/ffnet_54s-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_54s/releases/v0.51.0/ffnet_54s-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_54s/releases/v0.51.0/ffnet_54s-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_54s/releases/v0.51.0/ffnet_54s-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_54s/releases/v0.51.0/ffnet_54s-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[FFNet-54S on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/ffnet_54s)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/ffnet_54s) 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-54S on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/ffnet_54s) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: ffnet54S_dBBB_cityscapes_state_dict_quarts
- Input resolution: 2048x1024
- Number of output classes: 19
- Number of parameters: 18.0M
- Model size (float): 68.8 MB
- Model size (w8a8): 17.5 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| FFNet-54S | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.466 ms | 29 - 256 MB | NPU
| FFNet-54S | ONNX | float | Snapdragon® X2 Elite | 14.823 ms | 22 - 22 MB | NPU
| FFNet-54S | ONNX | float | Snapdragon® X Elite | 34.008 ms | 24 - 24 MB | NPU
| FFNet-54S | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 24.095 ms | 31 - 316 MB | NPU
| FFNet-54S | ONNX | float | Qualcomm® QCS8550 (Proxy) | 34.73 ms | 24 - 27 MB | NPU
| FFNet-54S | ONNX | float | Qualcomm® QCS9075 | 52.321 ms | 24 - 51 MB | NPU
| FFNet-54S | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 17.857 ms | 6 - 204 MB | NPU
| FFNet-54S | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 6.909 ms | 2 - 205 MB | NPU
| FFNet-54S | ONNX | w8a8 | Snapdragon® X2 Elite | 7.379 ms | 13 - 13 MB | NPU
| FFNet-54S | ONNX | w8a8 | Snapdragon® X Elite | 11.159 ms | 12 - 12 MB | NPU
| FFNet-54S | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 6.946 ms | 7 - 265 MB | NPU
| FFNet-54S | ONNX | w8a8 | Qualcomm® QCS6490 | 410.214 ms | 178 - 233 MB | CPU
| FFNet-54S | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 10.572 ms | 0 - 16 MB | NPU
| FFNet-54S | ONNX | w8a8 | Qualcomm® QCS9075 | 12.912 ms | 6 - 9 MB | NPU
| FFNet-54S | ONNX | w8a8 | Qualcomm® QCM6690 | 432.555 ms | 146 - 155 MB | CPU
| FFNet-54S | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 11.641 ms | 2 - 201 MB | NPU
| FFNet-54S | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 430.093 ms | 157 - 167 MB | CPU
| FFNet-54S | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.71 ms | 12 - 254 MB | NPU
| FFNet-54S | QNN_DLC | float | Snapdragon® X2 Elite | 15.895 ms | 24 - 24 MB | NPU
| FFNet-54S | QNN_DLC | float | Snapdragon® X Elite | 39.844 ms | 24 - 24 MB | NPU
| FFNet-54S | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 26.76 ms | 8 - 287 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 155.242 ms | 24 - 225 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 38.879 ms | 24 - 26 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® SA8775P | 53.669 ms | 24 - 225 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® QCS9075 | 66.395 ms | 24 - 52 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 78.284 ms | 11 - 281 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® SA7255P | 155.242 ms | 24 - 225 MB | NPU
| FFNet-54S | QNN_DLC | float | Qualcomm® SA8295P | 59.25 ms | 24 - 221 MB | NPU
| FFNet-54S | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 19.549 ms | 23 - 245 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 5.497 ms | 6 - 238 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 6.456 ms | 6 - 6 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Snapdragon® X Elite | 16.62 ms | 6 - 6 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 10.94 ms | 6 - 266 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 72.016 ms | 6 - 14 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 35.556 ms | 6 - 211 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 15.789 ms | 6 - 8 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® SA8775P | 16.274 ms | 6 - 213 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 19.124 ms | 6 - 14 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 143.124 ms | 6 - 244 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 22.56 ms | 6 - 265 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® SA7255P | 35.556 ms | 6 - 211 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Qualcomm® SA8295P | 21.316 ms | 6 - 215 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 7.482 ms | 6 - 225 MB | NPU
| FFNet-54S | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 19.831 ms | 6 - 229 MB | NPU
| FFNet-54S | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 13.699 ms | 2 - 261 MB | NPU
| FFNet-54S | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 26.887 ms | 1 - 329 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 155.512 ms | 3 - 224 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 39.138 ms | 2 - 5 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® SA8775P | 53.698 ms | 2 - 224 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® QCS9075 | 66.529 ms | 0 - 64 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 77.052 ms | 3 - 323 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® SA7255P | 155.512 ms | 3 - 224 MB | NPU
| FFNet-54S | TFLITE | float | Qualcomm® SA8295P | 59.289 ms | 0 - 216 MB | NPU
| FFNet-54S | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 19.671 ms | 0 - 239 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 3.171 ms | 1 - 232 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 5.925 ms | 1 - 258 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® QCS6490 | 55.719 ms | 0 - 27 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 22.846 ms | 1 - 204 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 8.267 ms | 1 - 12 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® SA8775P | 8.855 ms | 1 - 205 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® QCS9075 | 10.042 ms | 1 - 27 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® QCM6690 | 118.026 ms | 1 - 237 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 13.298 ms | 0 - 261 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® SA7255P | 22.846 ms | 1 - 204 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Qualcomm® SA8295P | 13.155 ms | 1 - 206 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 4.457 ms | 0 - 220 MB | NPU
| FFNet-54S | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 12.559 ms | 1 - 221 MB | NPU

## License
* The license for the original implementation of FFNet-54S 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).