Hugging Face's logo

qualcomm
/
FFNet-78S

Image Segmentation
PyTorch
LiteRT
ONNX
android
Model card Files Community

FFNet-78S: Optimized for Mobile Deployment

Semantic segmentation for automotive street scenes

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 model is an implementation of FFNet-78S found here.

This repository provides scripts to run FFNet-78S 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: ffnet78S_dBBB_cityscapes_state_dict_quarts
    • Input resolution: 2048x1024
    • Number of parameters: 27.5M
    • Number of output classes: 19
    • Model size (float): 105 MB
    • Model size (w8a8): 26.7 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
FFNet-78S float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 200.123 ms 2 - 74 MB NPU FFNet-78S.tflite
FFNet-78S float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN 1238.312 ms 24 - 34 MB NPU Use Export Script
FFNet-78S float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 67.199 ms 2 - 122 MB NPU FFNet-78S.tflite
FFNet-78S float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN 78.866 ms 24 - 77 MB NPU Use Export Script
FFNet-78S float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 49.602 ms 2 - 21 MB NPU FFNet-78S.tflite
FFNet-78S float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN 40.166 ms 24 - 26 MB NPU Use Export Script
FFNet-78S float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 68.935 ms 2 - 74 MB NPU FFNet-78S.tflite
FFNet-78S float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN 58.768 ms 24 - 36 MB NPU Use Export Script
FFNet-78S float SA7255P ADP Qualcomm® SA7255P TFLITE 200.123 ms 2 - 74 MB NPU FFNet-78S.tflite
FFNet-78S float SA7255P ADP Qualcomm® SA7255P QNN 1238.312 ms 24 - 34 MB NPU Use Export Script
FFNet-78S float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 49.849 ms 2 - 24 MB NPU FFNet-78S.tflite
FFNet-78S float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN 40.014 ms 24 - 26 MB NPU Use Export Script
FFNet-78S float SA8295P ADP Qualcomm® SA8295P TFLITE 77.09 ms 2 - 72 MB NPU FFNet-78S.tflite
FFNet-78S float SA8295P ADP Qualcomm® SA8295P QNN 65.831 ms 24 - 42 MB NPU Use Export Script
FFNet-78S float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 49.887 ms 2 - 26 MB NPU FFNet-78S.tflite
FFNet-78S float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN 40.024 ms 24 - 26 MB NPU Use Export Script
FFNet-78S float SA8775P ADP Qualcomm® SA8775P TFLITE 68.935 ms 2 - 74 MB NPU FFNet-78S.tflite
FFNet-78S float SA8775P ADP Qualcomm® SA8775P QNN 58.768 ms 24 - 36 MB NPU Use Export Script
FFNet-78S float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 49.623 ms 2 - 22 MB NPU FFNet-78S.tflite
FFNet-78S float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN 40.081 ms 24 - 47 MB NPU Use Export Script
FFNet-78S float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 41.493 ms 29 - 163 MB NPU FFNet-78S.onnx
FFNet-78S float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 34.026 ms 2 - 125 MB NPU FFNet-78S.tflite
FFNet-78S float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN 28.02 ms 24 - 76 MB NPU Use Export Script
FFNet-78S float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 29.94 ms 24 - 80 MB NPU FFNet-78S.onnx
FFNet-78S float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 28.695 ms 2 - 77 MB NPU FFNet-78S.tflite
FFNet-78S float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN 26.755 ms 24 - 86 MB NPU Use Export Script
FFNet-78S float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 26.697 ms 30 - 78 MB NPU FFNet-78S.onnx
FFNet-78S float Snapdragon X Elite CRD Snapdragon® X Elite QNN 41.57 ms 24 - 24 MB NPU Use Export Script
FFNet-78S float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 42.89 ms 30 - 30 MB NPU FFNet-78S.onnx
FFNet-78S w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 29.475 ms 1 - 50 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN 162.905 ms 6 - 16 MB NPU Use Export Script
FFNet-78S w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 12.252 ms 1 - 79 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN 20.713 ms 6 - 90 MB NPU Use Export Script
FFNet-78S w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 10.709 ms 1 - 14 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN 16.603 ms 6 - 9 MB NPU Use Export Script
FFNet-78S w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 11.141 ms 0 - 50 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN 17.089 ms 6 - 21 MB NPU Use Export Script
FFNet-78S w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) TFLITE 73.975 ms 1 - 77 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) QNN 84.703 ms 6 - 18 MB NPU Use Export Script
FFNet-78S w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 337.817 ms 1 - 12 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 29.475 ms 1 - 50 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 SA7255P ADP Qualcomm® SA7255P QNN 162.905 ms 6 - 16 MB NPU Use Export Script
FFNet-78S w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 10.737 ms 1 - 12 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN 16.585 ms 8 - 10 MB NPU Use Export Script
FFNet-78S w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 16.755 ms 1 - 52 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 SA8295P ADP Qualcomm® SA8295P QNN 23.588 ms 6 - 24 MB NPU Use Export Script
FFNet-78S w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 10.603 ms 1 - 14 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN 16.601 ms 6 - 8 MB NPU Use Export Script
FFNet-78S w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 11.141 ms 0 - 50 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 SA8775P ADP Qualcomm® SA8775P QNN 17.089 ms 6 - 21 MB NPU Use Export Script
FFNet-78S w8a8 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 10.557 ms 1 - 14 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN 16.512 ms 6 - 29 MB NPU Use Export Script
FFNet-78S w8a8 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 18.823 ms 2 - 94 MB NPU FFNet-78S.onnx
FFNet-78S w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 7.617 ms 1 - 75 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN 11.801 ms 6 - 91 MB NPU Use Export Script
FFNet-78S w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 14.797 ms 10 - 159 MB NPU FFNet-78S.onnx
FFNet-78S w8a8 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 7.431 ms 1 - 54 MB NPU FFNet-78S.tflite
FFNet-78S w8a8 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN 9.537 ms 6 - 79 MB NPU Use Export Script
FFNet-78S w8a8 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 12.725 ms 10 - 127 MB NPU FFNet-78S.onnx
FFNet-78S w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN 17.536 ms 6 - 6 MB NPU Use Export Script
FFNet-78S w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 22.144 ms 22 - 22 MB NPU FFNet-78S.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[ffnet-78s]"

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

Sign-in to Qualcomm® AI Hub 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.ffnet_78s.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.ffnet_78s.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.ffnet_78s.export

Profiling Results
------------------------------------------------------------
FFNet-78S
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 200.1                                
Estimated peak memory usage (MB): [2, 74]                              
Total # Ops                     : 151                                  
Compute Unit(s)                 : npu (151 ops) gpu (0 ops) cpu (0 ops)

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.ffnet_78s 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.

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. 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.ffnet_78s.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.ffnet_78s.demo -- --on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on FFNet-78S's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of FFNet-78S can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month
51
Inference Providers NEW
Image Segmentation
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support