library_name: pytorch
license: other
tags:
- real_time
- android
pipeline_tag: image-segmentation
PidNet: Optimized for Mobile Deployment
Segment images or video by class in real-time on device
PIDNet (Proportional-Integral-Derivative Network) is a real-time semantic segmentation model based on PID controllers
This model is an implementation of PidNet found here.
This repository provides scripts to run PidNet on Qualcomm® devices. More details on model performance across various devices, can be found here.
Model Details
- Model Type: Model_use_case.semantic_segmentation
- Model Stats:
- Model checkpoint: PIDNet_S_Cityscapes_val.pt
- Inference latency: RealTime
- Input resolution: 1024x2048
- Number of output classes: 19
- Number of parameters: 8.06M
- Model size (float): 29.1 MB
- Model size (w8a8): 8.02 MB
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| PidNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 136.7 ms | 0 - 57 MB | NPU | PidNet.tflite |
| PidNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 115.422 ms | 24 - 94 MB | NPU | PidNet.dlc |
| PidNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 58.67 ms | 2 - 70 MB | NPU | PidNet.tflite |
| PidNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 69.219 ms | 24 - 110 MB | NPU | PidNet.dlc |
| PidNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 48.184 ms | 2 - 26 MB | NPU | PidNet.tflite |
| PidNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 36.215 ms | 24 - 50 MB | NPU | PidNet.dlc |
| PidNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 29.548 ms | 24 - 85 MB | NPU | PidNet.onnx.zip |
| PidNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 58.213 ms | 2 - 59 MB | NPU | PidNet.tflite |
| PidNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 44.393 ms | 24 - 93 MB | NPU | PidNet.dlc |
| PidNet | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 136.7 ms | 0 - 57 MB | NPU | PidNet.tflite |
| PidNet | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 115.422 ms | 24 - 94 MB | NPU | PidNet.dlc |
| PidNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 48.699 ms | 2 - 31 MB | NPU | PidNet.tflite |
| PidNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 35.917 ms | 24 - 51 MB | NPU | PidNet.dlc |
| PidNet | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 65.89 ms | 2 - 61 MB | NPU | PidNet.tflite |
| PidNet | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 50.378 ms | 23 - 102 MB | NPU | PidNet.dlc |
| PidNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 48.605 ms | 2 - 33 MB | NPU | PidNet.tflite |
| PidNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 36.219 ms | 24 - 55 MB | NPU | PidNet.dlc |
| PidNet | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 58.213 ms | 2 - 59 MB | NPU | PidNet.tflite |
| PidNet | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 44.393 ms | 24 - 93 MB | NPU | PidNet.dlc |
| PidNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 31.461 ms | 1 - 71 MB | NPU | PidNet.tflite |
| PidNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 24.998 ms | 24 - 97 MB | NPU | PidNet.dlc |
| PidNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 20.064 ms | 30 - 105 MB | NPU | PidNet.onnx.zip |
| PidNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 25.056 ms | 0 - 61 MB | NPU | PidNet.tflite |
| PidNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 18.638 ms | 24 - 106 MB | NPU | PidNet.dlc |
| PidNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 16.172 ms | 7 - 82 MB | NPU | PidNet.onnx.zip |
| PidNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 21.762 ms | 1 - 61 MB | NPU | PidNet.tflite |
| PidNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 13.717 ms | 24 - 122 MB | NPU | PidNet.dlc |
| PidNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 11.471 ms | 30 - 147 MB | NPU | PidNet.onnx.zip |
| PidNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 36.936 ms | 24 - 24 MB | NPU | PidNet.dlc |
| PidNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 29.762 ms | 24 - 24 MB | NPU | PidNet.onnx.zip |
| PidNet | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | TFLITE | 172.215 ms | 2 - 72 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 390.44 ms | 195 - 216 MB | CPU | PidNet.onnx.zip |
| PidNet | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 103.389 ms | 0 - 46 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 125.244 ms | 6 - 70 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 52.636 ms | 1 - 56 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 77.466 ms | 6 - 85 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 52.448 ms | 0 - 25 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 64.955 ms | 6 - 30 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 61.782 ms | 91 - 113 MB | NPU | PidNet.onnx.zip |
| PidNet | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 53.101 ms | 1 - 45 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 65.749 ms | 6 - 69 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 354.579 ms | 190 - 201 MB | CPU | PidNet.onnx.zip |
| PidNet | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 103.389 ms | 0 - 46 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 125.244 ms | 6 - 70 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 52.584 ms | 0 - 20 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 64.764 ms | 6 - 31 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 61.026 ms | 1 - 49 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 73.693 ms | 6 - 71 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 52.222 ms | 0 - 20 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 65.107 ms | 6 - 24 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 53.101 ms | 1 - 45 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 65.749 ms | 6 - 69 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 39.546 ms | 1 - 57 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 48.901 ms | 6 - 84 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 48.515 ms | 105 - 168 MB | NPU | PidNet.onnx.zip |
| PidNet | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 38.702 ms | 1 - 50 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 41.3 ms | 6 - 82 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 42.726 ms | 101 - 160 MB | NPU | PidNet.onnx.zip |
| PidNet | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | TFLITE | 69.393 ms | 1 - 53 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 327.861 ms | 191 - 208 MB | CPU | PidNet.onnx.zip |
| PidNet | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 24.511 ms | 1 - 50 MB | NPU | PidNet.tflite |
| PidNet | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 43.463 ms | 6 - 101 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 43.23 ms | 105 - 169 MB | NPU | PidNet.onnx.zip |
| PidNet | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 68.103 ms | 21 - 21 MB | NPU | PidNet.dlc |
| PidNet | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 62.524 ms | 131 - 131 MB | NPU | PidNet.onnx.zip |
Installation
Install the package via pip:
pip install qai-hub-models
Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device
Sign-in to Qualcomm® AI Hub Workbench 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.
qai-hub configure --api_token API_TOKEN
Navigate to 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.
python -m qai_hub_models.models.pidnet.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.pidnet.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.
python -m qai_hub_models.models.pidnet.export
How does this work?
This export script leverages Qualcomm® AI Hub 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.
import torch
import qai_hub as hub
from qai_hub_models.models.pidnet import Model
# Load the model
torch_model = Model.from_pretrained()
# Device
device = hub.Device("Samsung Galaxy S25")
# 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.
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.
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 Workbench. Sign up for access.
Run demo on a cloud-hosted device
You can also run the demo on-device.
python -m qai_hub_models.models.pidnet.demo --eval-mode 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.pidnet.demo -- --eval-mode on-device
Deploying compiled model to Android
The models can be deployed using multiple runtimes:
TensorFlow Lite (
.tfliteexport): This tutorial provides a guide to deploy the .tflite model in an Android application.QNN (
.soexport ): This sample app provides instructions on how to use the.soshared library in an Android application.
View on Qualcomm® AI Hub
Get more details on PidNet's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of PidNet can be found here.
- The license for the compiled assets for on-device deployment can be found here
References
- PIDNet A Real-time Semantic Segmentation Network Inspired from PID Controller Segmentation of Road Scenes
- Source Model Implementation
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.