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

# Segformer-Base: Optimized for Qualcomm Devices

Segformer Base is a machine learning model that predicts masks and classes of objects in an image.

This is based on the implementation of Segformer-Base found [here](https://github.com/huggingface/transformers/tree/main/src/transformers/models/segformer).
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/segformer_base) 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/segformer_base/releases/v0.50.2/segformer_base-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/segformer_base/releases/v0.50.2/segformer_base-onnx-w8a16.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/segformer_base/releases/v0.50.2/segformer_base-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/segformer_base/releases/v0.50.2/segformer_base-qnn_dlc-float.zip)
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.19.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/segformer_base/releases/v0.50.2/segformer_base-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.19.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/segformer_base/releases/v0.50.2/segformer_base-tflite-w8a8.zip)

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


### 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/segformer_base) 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 [Segformer-Base on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/segformer_base) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: nvidia/segformer-b0-finetuned-ade-512-512
- Input resolution: 512x512
- Number of output classes: 150
- Number of parameters: 3.75M
- Model size (float): 14.4 MB
- Model size (w8a16): 4.57 MB
- Model size (w8a8): 3.90 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| Segformer-Base | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 74.11 ms | 24 - 216 MB | NPU
| Segformer-Base | ONNX | float | Snapdragon® X2 Elite | 72.551 ms | 34 - 34 MB | NPU
| Segformer-Base | ONNX | float | Snapdragon® X Elite | 112.531 ms | 33 - 33 MB | NPU
| Segformer-Base | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 82.136 ms | 25 - 256 MB | NPU
| Segformer-Base | ONNX | float | Qualcomm® QCS8550 (Proxy) | 108.436 ms | 19 - 28 MB | NPU
| Segformer-Base | ONNX | float | Qualcomm® QCS9075 | 113.253 ms | 23 - 26 MB | NPU
| Segformer-Base | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 74.201 ms | 23 - 214 MB | NPU
| Segformer-Base | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 6.779 ms | 14 - 221 MB | NPU
| Segformer-Base | ONNX | w8a16 | Snapdragon® X2 Elite | 6.819 ms | 13 - 13 MB | NPU
| Segformer-Base | ONNX | w8a16 | Snapdragon® X Elite | 15.321 ms | 18 - 18 MB | NPU
| Segformer-Base | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 10.353 ms | 12 - 252 MB | NPU
| Segformer-Base | ONNX | w8a16 | Qualcomm® QCS6490 | 742.54 ms | 385 - 391 MB | CPU
| Segformer-Base | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 14.858 ms | 9 - 16 MB | NPU
| Segformer-Base | ONNX | w8a16 | Qualcomm® QCS9075 | 20.798 ms | 12 - 15 MB | NPU
| Segformer-Base | ONNX | w8a16 | Qualcomm® QCM6690 | 353.439 ms | 333 - 344 MB | CPU
| Segformer-Base | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 8.554 ms | 13 - 217 MB | NPU
| Segformer-Base | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 318.687 ms | 366 - 378 MB | CPU
| Segformer-Base | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 4.572 ms | 6 - 202 MB | NPU
| Segformer-Base | ONNX | w8a8 | Snapdragon® X2 Elite | 4.602 ms | 4 - 4 MB | NPU
| Segformer-Base | ONNX | w8a8 | Snapdragon® X Elite | 11.675 ms | 9 - 9 MB | NPU
| Segformer-Base | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 7.597 ms | 6 - 231 MB | NPU
| Segformer-Base | ONNX | w8a8 | Qualcomm® QCS6490 | 277.259 ms | 194 - 202 MB | CPU
| Segformer-Base | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 10.978 ms | 2 - 13 MB | NPU
| Segformer-Base | ONNX | w8a8 | Qualcomm® QCS9075 | 11.572 ms | 7 - 10 MB | NPU
| Segformer-Base | ONNX | w8a8 | Qualcomm® QCM6690 | 175.328 ms | 195 - 207 MB | CPU
| Segformer-Base | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 5.575 ms | 8 - 204 MB | NPU
| Segformer-Base | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 157.633 ms | 128 - 139 MB | CPU
| Segformer-Base | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 73.829 ms | 3 - 194 MB | NPU
| Segformer-Base | QNN_DLC | float | Snapdragon® X2 Elite | 73.271 ms | 3 - 3 MB | NPU
| Segformer-Base | QNN_DLC | float | Snapdragon® X Elite | 114.593 ms | 3 - 3 MB | NPU
| Segformer-Base | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 83.847 ms | 0 - 226 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 210.561 ms | 0 - 185 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 110.179 ms | 3 - 78 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® SA8775P | 112.414 ms | 1 - 183 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® QCS9075 | 113.595 ms | 3 - 17 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 122.184 ms | 2 - 223 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® SA7255P | 210.561 ms | 0 - 185 MB | NPU
| Segformer-Base | QNN_DLC | float | Qualcomm® SA8295P | 122.632 ms | 3 - 184 MB | NPU
| Segformer-Base | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 74.571 ms | 3 - 196 MB | NPU
| Segformer-Base | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 73.979 ms | 16 - 211 MB | NPU
| Segformer-Base | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 82.7 ms | 8 - 235 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 210.505 ms | 10 - 197 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 110.047 ms | 9 - 12 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® SA8775P | 112.394 ms | 10 - 194 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® QCS9075 | 112.427 ms | 8 - 30 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 123.061 ms | 0 - 222 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® SA7255P | 210.505 ms | 10 - 197 MB | NPU
| Segformer-Base | TFLITE | float | Qualcomm® SA8295P | 122.654 ms | 9 - 194 MB | NPU
| Segformer-Base | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 74.914 ms | 9 - 194 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 4.401 ms | 2 - 185 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 7.549 ms | 2 - 209 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® QCS6490 | 126.202 ms | 15 - 50 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 19.666 ms | 2 - 177 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 10.884 ms | 2 - 6 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® SA8775P | 11.475 ms | 2 - 178 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® QCS9075 | 11.357 ms | 0 - 10 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® QCM6690 | 95.908 ms | 13 - 176 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 14.871 ms | 2 - 210 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® SA7255P | 19.666 ms | 2 - 177 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Qualcomm® SA8295P | 13.593 ms | 2 - 182 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 5.53 ms | 0 - 173 MB | NPU
| Segformer-Base | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 39.272 ms | 15 - 63 MB | NPU

## License
* The license for the original implementation of Segformer-Base can be found
  [here](https://github.com/huggingface/transformers/blob/main/LICENSE).

## References
* [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203)
* [Source Model Implementation](https://github.com/huggingface/transformers/tree/main/src/transformers/models/segformer)

## 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).