---
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_78s/web-assets/model_demo.png)

# FFNet-78S: Optimized for Qualcomm Devices

FFNet-78S 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-78S 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_78s) 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_78s/releases/v0.51.0/ffnet_78s-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_78s/releases/v0.51.0/ffnet_78s-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_78s/releases/v0.51.0/ffnet_78s-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_78s/releases/v0.51.0/ffnet_78s-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_78s/releases/v0.51.0/ffnet_78s-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_78s/releases/v0.51.0/ffnet_78s-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[FFNet-78S on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/ffnet_78s)**.


### 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_78s) 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-78S on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/ffnet_78s) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: ffnet78S_dBBB_cityscapes_state_dict_quarts
- Input resolution: 2048x1024
- Number of output classes: 19
- Number of parameters: 27.5M
- Model size (float): 105 MB
- Model size (w8a8): 26.7 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| FFNet-78S | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 17.173 ms | 29 - 258 MB | NPU
| FFNet-78S | ONNX | float | Snapdragon® X2 Elite | 17.872 ms | 30 - 30 MB | NPU
| FFNet-78S | ONNX | float | Snapdragon® X Elite | 37.856 ms | 30 - 30 MB | NPU
| FFNet-78S | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 27.043 ms | 0 - 301 MB | NPU
| FFNet-78S | ONNX | float | Qualcomm® QCS8550 (Proxy) | 38.338 ms | 24 - 27 MB | NPU
| FFNet-78S | ONNX | float | Qualcomm® QCS9075 | 59.331 ms | 24 - 51 MB | NPU
| FFNet-78S | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 20.224 ms | 6 - 209 MB | NPU
| FFNet-78S | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 7.282 ms | 0 - 219 MB | NPU
| FFNet-78S | ONNX | w8a8 | Snapdragon® X2 Elite | 7.925 ms | 22 - 22 MB | NPU
| FFNet-78S | ONNX | w8a8 | Snapdragon® X Elite | 14.884 ms | 21 - 21 MB | NPU
| FFNet-78S | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 11.938 ms | 7 - 295 MB | NPU
| FFNet-78S | ONNX | w8a8 | Qualcomm® QCS6490 | 507.691 ms | 162 - 230 MB | CPU
| FFNet-78S | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 15.421 ms | 0 - 102 MB | NPU
| FFNet-78S | ONNX | w8a8 | Qualcomm® QCS9075 | 14.482 ms | 6 - 9 MB | NPU
| FFNet-78S | ONNX | w8a8 | Qualcomm® QCM6690 | 533.377 ms | 139 - 148 MB | CPU
| FFNet-78S | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 8.78 ms | 1 - 209 MB | NPU
| FFNet-78S | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 532.845 ms | 148 - 158 MB | CPU
| FFNet-78S | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 15.463 ms | 24 - 273 MB | NPU
| FFNet-78S | QNN_DLC | float | Snapdragon® X2 Elite | 17.953 ms | 24 - 24 MB | NPU
| FFNet-78S | QNN_DLC | float | Snapdragon® X Elite | 44.25 ms | 24 - 24 MB | NPU
| FFNet-78S | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 29.407 ms | 19 - 315 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 187.95 ms | 24 - 233 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 42.6 ms | 24 - 26 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® SA8775P | 60.845 ms | 24 - 234 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® QCS9075 | 73.42 ms | 24 - 52 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 86.396 ms | 4 - 285 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® SA7255P | 187.95 ms | 24 - 233 MB | NPU
| FFNet-78S | QNN_DLC | float | Qualcomm® SA8295P | 66.496 ms | 24 - 228 MB | NPU
| FFNet-78S | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 21.845 ms | 24 - 254 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 5.884 ms | 6 - 250 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 7.009 ms | 6 - 6 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Snapdragon® X Elite | 17.604 ms | 6 - 6 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 11.609 ms | 6 - 291 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 72.281 ms | 8 - 16 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 39.065 ms | 6 - 225 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 16.781 ms | 6 - 8 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® SA8775P | 17.196 ms | 6 - 226 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 20.011 ms | 6 - 14 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 165.642 ms | 6 - 256 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 23.657 ms | 6 - 293 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® SA7255P | 39.065 ms | 6 - 225 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Qualcomm® SA8295P | 23.328 ms | 6 - 227 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 8.06 ms | 6 - 238 MB | NPU
| FFNet-78S | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 21.825 ms | 6 - 242 MB | NPU
| FFNet-78S | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 15.483 ms | 2 - 281 MB | NPU
| FFNet-78S | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 29.449 ms | 2 - 378 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 187.958 ms | 3 - 245 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 42.887 ms | 2 - 5 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® SA8775P | 61.023 ms | 2 - 245 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® QCS9075 | 73.561 ms | 0 - 82 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 88.15 ms | 3 - 362 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® SA7255P | 187.958 ms | 3 - 245 MB | NPU
| FFNet-78S | TFLITE | float | Qualcomm® SA8295P | 66.475 ms | 2 - 240 MB | NPU
| FFNet-78S | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 21.775 ms | 2 - 264 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 3.578 ms | 0 - 242 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 6.465 ms | 1 - 290 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® QCS6490 | 57.526 ms | 1 - 36 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 26.245 ms | 1 - 218 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 9.068 ms | 1 - 3 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® SA8775P | 9.723 ms | 1 - 219 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® QCS9075 | 11.161 ms | 1 - 36 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® QCM6690 | 139.037 ms | 1 - 248 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 18.45 ms | 1 - 291 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® SA7255P | 26.245 ms | 1 - 218 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Qualcomm® SA8295P | 14.886 ms | 1 - 221 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 5.04 ms | 1 - 231 MB | NPU
| FFNet-78S | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 14.574 ms | 1 - 234 MB | NPU

## License
* The license for the original implementation of FFNet-78S 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).