Upload README.md with huggingface_hub

84b8214 verified over 1 year ago

16.6 kB

---
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/deeplabv3_plus_mobilenet/web-assets/model_demo.png)

# DeepLabV3-Plus-MobileNet: Optimized for Mobile Deployment
## Deep Convolutional Neural Network model for semantic segmentation


DeepLabV3 is designed for semantic segmentation at multiple scales, trained on the various datasets. It uses MobileNet as a backbone.

This model is an implementation of DeepLabV3-Plus-MobileNet found [here](https://github.com/jfzhang95/pytorch-deeplab-xception).


This repository provides scripts to run DeepLabV3-Plus-MobileNet on Qualcomm® devices.
More details on model performance across various devices, can be found
[here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet).


### Model Details

- **Model Type:** Model_use_case.semantic_segmentation
- **Model Stats:**
  - Model checkpoint: VOC2012
  - Input resolution: 513x513
  - Number of parameters: 5.80M
  - Number of output classes: 21
  - Model size (float): 22.2 MB
  - Model size (w8a8): 6.04 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|---|---|---|---|---|---|---|---|---|
| DeepLabV3-Plus-MobileNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 62.823 ms | 0 - 26 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 58.604 ms | 3 - 13 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 19.276 ms | 0 - 44 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 21.781 ms | 3 - 46 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 12.959 ms | 0 - 16 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 11.371 ms | 3 - 5 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 19.64 ms | 0 - 28 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 17.468 ms | 0 - 15 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 62.823 ms | 0 - 26 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 58.604 ms | 3 - 13 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 13.1 ms | 0 - 12 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 11.393 ms | 3 - 6 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 22.247 ms | 0 - 27 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 21.14 ms | 0 - 18 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 13.034 ms | 0 - 12 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 11.364 ms | 4 - 5 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 19.64 ms | 0 - 28 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 17.468 ms | 0 - 15 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 13.069 ms | 0 - 7 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 11.286 ms | 3 - 15 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 10.661 ms | 1 - 40 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) |
| DeepLabV3-Plus-MobileNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 9.114 ms | 0 - 47 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 7.825 ms | 3 - 43 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 7.512 ms | 1 - 43 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) |
| DeepLabV3-Plus-MobileNet | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 8.61 ms | 0 - 30 MB | NPU | [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) |
| DeepLabV3-Plus-MobileNet | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 7.555 ms | 3 - 40 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 7.463 ms | 3 - 36 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) |
| DeepLabV3-Plus-MobileNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 11.973 ms | 3 - 3 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 11.837 ms | 10 - 10 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) |
| DeepLabV3-Plus-MobileNet | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 21.221 ms | 2 - 11 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 11.951 ms | 2 - 51 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 8.299 ms | 2 - 4 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 8.893 ms | 2 - 12 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN | 43.959 ms | 2 - 15 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN | 21.221 ms | 2 - 11 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 8.301 ms | 2 - 4 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN | 12.554 ms | 2 - 20 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 8.319 ms | 2 - 4 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN | 8.893 ms | 2 - 12 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 8.261 ms | 2 - 16 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 74.094 ms | 93 - 141 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) |
| DeepLabV3-Plus-MobileNet | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 5.907 ms | 2 - 52 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 58.434 ms | 97 - 247 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) |
| DeepLabV3-Plus-MobileNet | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 4.979 ms | 2 - 45 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 58.233 ms | 84 - 227 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) |
| DeepLabV3-Plus-MobileNet | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 8.904 ms | 2 - 2 MB | NPU | Use Export Script |
| DeepLabV3-Plus-MobileNet | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 72.494 ms | 133 - 133 MB | NPU | [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) |
## Installation
Install the package via pip:
```bash
pip install qai-hub-models
```
## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
With this API token, you can configure your client to run models on the cloud
hosted devices.
```bash
qai-hub configure --api_token API_TOKEN
```
Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.
## Demo off target
The package contains a simple end-to-end demo that downloads pre-trained
weights and runs this model on a sample input.
```bash
python -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo
```
The above demo runs a reference implementation of pre-processing, model
inference, and post processing.
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo
```
### Run model on a cloud-hosted device
In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
device. This script does the following:
* Performance check on-device on a cloud-hosted device
* Downloads compiled assets that can be deployed on-device for Android.
* Accuracy check between PyTorch and on-device outputs.
```bash
python -m qai_hub_models.models.deeplabv3_plus_mobilenet.export
```
```
Profiling Results
------------------------------------------------------------
DeepLabV3-Plus-MobileNet
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 62.8                                 
Estimated peak memory usage (MB): [0, 26]                              
Total # Ops                     : 101                                  
Compute Unit(s)                 : npu (101 ops) gpu (0 ops) cpu (0 ops)
```
## How does this work?
This [export script](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet/qai_hub_models/models/DeepLabV3-Plus-MobileNet/export.py)
leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
on-device. Lets go through each step below in detail:
Step 1: **Compile model for on-device deployment**
To compile a PyTorch model for on-device deployment, we first trace the model
in memory using the `jit.trace` and then call the `submit_compile_job` API.
```python
import torch

import qai_hub as hub
from qai_hub_models.models.deeplabv3_plus_mobilenet import Model
# Load the model
torch_model = Model.from_pretrained()
# Device
device = hub.Device("Samsung Galaxy S24")
# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)
# Get target model to run on-device
target_model = compile_job.get_target_model()

```
Step 2: **Performance profiling on cloud-hosted device**
After compiling models from step 1. Models can be profiled model on-device using the
`target_model`. Note that this scripts runs the model on a device automatically
provisioned in the cloud.  Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
```python
profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)
        
```
Step 3: **Verify on-device accuracy**

To verify the accuracy of the model on-device, you can run on-device inference
on sample input data on the same cloud hosted device.
```python
input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()
```
With the output of the model, you can compute like PSNR, relative errors or
spot check the output with expected output.

**Note**: This on-device profiling and inference requires access to Qualcomm®
AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).



## Run demo on a cloud-hosted device

You can also run the demo on-device.

```bash
python -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo --on-device
```

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo -- --on-device
```


## Deploying compiled model to Android


The models can be deployed using multiple runtimes:
- TensorFlow Lite (`.tflite` export): [This
  tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
  guide to deploy the .tflite model in an Android application.


- QNN (`.so` export ): This [sample
  app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
provides instructions on how to use the `.so` shared library  in an Android application.
## View on Qualcomm® AI Hub
Get more details on DeepLabV3-Plus-MobileNet's performance across various devices [here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet).
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
## License
* The license for the original implementation of DeepLabV3-Plus-MobileNet can be found
  [here](https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/LICENSE).
* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
## References
* [Rethinking Atrous Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1706.05587)
* [Source Model Implementation](https://github.com/jfzhang95/pytorch-deeplab-xception)
## 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).

	---
	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/deeplabv3_plus_mobilenet/web-assets/model_demo.png)

	# DeepLabV3-Plus-MobileNet: Optimized for Mobile Deployment
	## Deep Convolutional Neural Network model for semantic segmentation


	DeepLabV3 is designed for semantic segmentation at multiple scales, trained on the various datasets. It uses MobileNet as a backbone.

	This model is an implementation of DeepLabV3-Plus-MobileNet found [here](https://github.com/jfzhang95/pytorch-deeplab-xception).


	This repository provides scripts to run DeepLabV3-Plus-MobileNet on Qualcomm® devices.
	More details on model performance across various devices, can be found
	[here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet).


	### Model Details

	- Model Type: Model_use_case.semantic_segmentation
	- Model Stats:
	- Model checkpoint: VOC2012
	- Input resolution: 513x513
	- Number of parameters: 5.80M
	- Number of output classes: 21
	- Model size (float): 22.2 MB
	- Model size (w8a8): 6.04 MB

	\| Model \| Precision \| Device \| Chipset \| Target Runtime \| Inference Time (ms) \| Peak Memory Range (MB) \| Primary Compute Unit \| Target Model
	\|---\|---\|---\|---\|---\|---\|---\|---\|---\|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| TFLITE \| 62.823 ms \| 0 - 26 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| QNN \| 58.604 ms \| 3 - 13 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| TFLITE \| 19.276 ms \| 0 - 44 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| QNN \| 21.781 ms \| 3 - 46 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| TFLITE \| 12.959 ms \| 0 - 16 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| QNN \| 11.371 ms \| 3 - 5 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| TFLITE \| 19.64 ms \| 0 - 28 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| QNN \| 17.468 ms \| 0 - 15 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA7255P ADP \| Qualcomm® SA7255P \| TFLITE \| 62.823 ms \| 0 - 26 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA7255P ADP \| Qualcomm® SA7255P \| QNN \| 58.604 ms \| 3 - 13 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| TFLITE \| 13.1 ms \| 0 - 12 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| QNN \| 11.393 ms \| 3 - 6 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8295P ADP \| Qualcomm® SA8295P \| TFLITE \| 22.247 ms \| 0 - 27 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8295P ADP \| Qualcomm® SA8295P \| QNN \| 21.14 ms \| 0 - 18 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| TFLITE \| 13.034 ms \| 0 - 12 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| QNN \| 11.364 ms \| 4 - 5 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8775P ADP \| Qualcomm® SA8775P \| TFLITE \| 19.64 ms \| 0 - 28 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| SA8775P ADP \| Qualcomm® SA8775P \| QNN \| 17.468 ms \| 0 - 15 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| TFLITE \| 13.069 ms \| 0 - 7 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| QNN \| 11.286 ms \| 3 - 15 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| ONNX \| 10.661 ms \| 1 - 40 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| TFLITE \| 9.114 ms \| 0 - 47 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| QNN \| 7.825 ms \| 3 - 43 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| ONNX \| 7.512 ms \| 1 - 43 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| TFLITE \| 8.61 ms \| 0 - 30 MB \| NPU \| [DeepLabV3-Plus-MobileNet.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.tflite) \|
	\| DeepLabV3-Plus-MobileNet \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| QNN \| 7.555 ms \| 3 - 40 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| ONNX \| 7.463 ms \| 3 - 36 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| float \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| QNN \| 11.973 ms \| 3 - 3 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| float \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| ONNX \| 11.837 ms \| 10 - 10 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| QNN \| 21.221 ms \| 2 - 11 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| QNN \| 11.951 ms \| 2 - 51 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| QNN \| 8.299 ms \| 2 - 4 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| QNN \| 8.893 ms \| 2 - 12 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| RB3 Gen 2 (Proxy) \| Qualcomm® QCS6490 (Proxy) \| QNN \| 43.959 ms \| 2 - 15 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| SA7255P ADP \| Qualcomm® SA7255P \| QNN \| 21.221 ms \| 2 - 11 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| QNN \| 8.301 ms \| 2 - 4 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| SA8295P ADP \| Qualcomm® SA8295P \| QNN \| 12.554 ms \| 2 - 20 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| QNN \| 8.319 ms \| 2 - 4 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| SA8775P ADP \| Qualcomm® SA8775P \| QNN \| 8.893 ms \| 2 - 12 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| QNN \| 8.261 ms \| 2 - 16 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| ONNX \| 74.094 ms \| 93 - 141 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| QNN \| 5.907 ms \| 2 - 52 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| ONNX \| 58.434 ms \| 97 - 247 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| QNN \| 4.979 ms \| 2 - 45 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| ONNX \| 58.233 ms \| 84 - 227 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| QNN \| 8.904 ms \| 2 - 2 MB \| NPU \| Use Export Script \|
	\| DeepLabV3-Plus-MobileNet \| w8a16 \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| ONNX \| 72.494 ms \| 133 - 133 MB \| NPU \| [DeepLabV3-Plus-MobileNet.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet/blob/main/DeepLabV3-Plus-MobileNet_w8a16.onnx) \|




	## Installation


	Install the package via pip:
	```bash
	pip install qai-hub-models
	```


	## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

	Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
	Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.

	With this API token, you can configure your client to run models on the cloud
	hosted devices.
	```bash
	qai-hub configure --api_token API_TOKEN
	```
	Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.



	## Demo off target

	The package contains a simple end-to-end demo that downloads pre-trained
	weights and runs this model on a sample input.

	```bash
	python -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo
	```

	The above demo runs a reference implementation of pre-processing, model
	inference, and post processing.

	NOTE: If you want running in a Jupyter Notebook or Google Colab like
	environment, please add the following to your cell (instead of the above).
	```
	%run -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo
	```


	### Run model on a cloud-hosted device

	In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
	device. This script does the following:
	* Performance check on-device on a cloud-hosted device
	* Downloads compiled assets that can be deployed on-device for Android.
	* Accuracy check between PyTorch and on-device outputs.

	```bash
	python -m qai_hub_models.models.deeplabv3_plus_mobilenet.export
	```
	```
	Profiling Results
	------------------------------------------------------------
	DeepLabV3-Plus-MobileNet
	Device : cs_8275 (ANDROID 14)
	Runtime : TFLITE
	Estimated inference time (ms) : 62.8
	Estimated peak memory usage (MB): [0, 26]
	Total # Ops : 101
	Compute Unit(s) : npu (101 ops) gpu (0 ops) cpu (0 ops)
	```


	## How does this work?

	This [export script](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet/qai_hub_models/models/DeepLabV3-Plus-MobileNet/export.py)
	leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
	on-device. Lets go through each step below in detail:

	Step 1: Compile model for on-device deployment

	To compile a PyTorch model for on-device deployment, we first trace the model
	in memory using the `jit.trace` and then call the `submit_compile_job` API.

	```python
	import torch

	import qai_hub as hub
	from qai_hub_models.models.deeplabv3_plus_mobilenet import Model

	# Load the model
	torch_model = Model.from_pretrained()

	# Device
	device = hub.Device("Samsung Galaxy S24")

	# Trace model
	input_shape = torch_model.get_input_spec()
	sample_inputs = torch_model.sample_inputs()

	pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

	# Compile model on a specific device
	compile_job = hub.submit_compile_job(
	model=pt_model,
	device=device,
	input_specs=torch_model.get_input_spec(),
	)

	# Get target model to run on-device
	target_model = compile_job.get_target_model()

	```


	Step 2: Performance profiling on cloud-hosted device

	After compiling models from step 1. Models can be profiled model on-device using the
	`target_model`. Note that this scripts runs the model on a device automatically
	provisioned in the cloud. Once the job is submitted, you can navigate to a
	provided job URL to view a variety of on-device performance metrics.
	```python
	profile_job = hub.submit_profile_job(
	model=target_model,
	device=device,
	)

	```

	Step 3: Verify on-device accuracy

	To verify the accuracy of the model on-device, you can run on-device inference
	on sample input data on the same cloud hosted device.
	```python
	input_data = torch_model.sample_inputs()
	inference_job = hub.submit_inference_job(
	model=target_model,
	device=device,
	inputs=input_data,
	)
	on_device_output = inference_job.download_output_data()

	```
	With the output of the model, you can compute like PSNR, relative errors or
	spot check the output with expected output.

	Note: This on-device profiling and inference requires access to Qualcomm®
	AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).



	## Run demo on a cloud-hosted device

	You can also run the demo on-device.

	```bash
	python -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo --on-device
	```

	NOTE: If you want running in a Jupyter Notebook or Google Colab like
	environment, please add the following to your cell (instead of the above).
	```
	%run -m qai_hub_models.models.deeplabv3_plus_mobilenet.demo -- --on-device
	```


	## Deploying compiled model to Android


	The models can be deployed using multiple runtimes:
	- TensorFlow Lite (`.tflite` export): [This
	tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
	guide to deploy the .tflite model in an Android application.


	- QNN (`.so` export ): This [sample
	app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
	provides instructions on how to use the `.so` shared library in an Android application.


	## View on Qualcomm® AI Hub
	Get more details on DeepLabV3-Plus-MobileNet's performance across various devices [here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet).
	Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


	## License
	* The license for the original implementation of DeepLabV3-Plus-MobileNet can be found
	[here](https://github.com/jfzhang95/pytorch-deeplab-xception/blob/master/LICENSE).
	* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)



	## References
	* [Rethinking Atrous Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1706.05587)
	* [Source Model Implementation](https://github.com/jfzhang95/pytorch-deeplab-xception)



	## 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).