Hugging Face's logo

qualcomm
/
GoogLeNet

Image Classification
PyTorch
LiteRT
android
Model card Files
xet
Community

GoogLeNet: Optimized for Mobile Deployment

Imagenet classifier and general purpose backbone

GoogLeNet is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.

This model is an implementation of GoogLeNet found here.

This repository provides scripts to run GoogLeNet on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.image_classification
  • Model Stats:
    • Model checkpoint: Imagenet
    • Input resolution: 224x224
    • Number of parameters: 6.62M
    • Model size (float): 25.3 MB
    • Model size (w8a8): 6.54 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
GoogLeNet float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 5.022 ms 0 - 127 MB NPU GoogLeNet.tflite
GoogLeNet float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 5.058 ms 1 - 122 MB NPU GoogLeNet.dlc
GoogLeNet float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1.764 ms 0 - 151 MB NPU GoogLeNet.tflite
GoogLeNet float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 1.756 ms 1 - 143 MB NPU GoogLeNet.dlc
GoogLeNet float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.839 ms 0 - 3 MB NPU GoogLeNet.tflite
GoogLeNet float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.838 ms 1 - 3 MB NPU GoogLeNet.dlc
GoogLeNet float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 1.098 ms 0 - 16 MB NPU GoogLeNet.onnx.zip
GoogLeNet float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1.584 ms 0 - 128 MB NPU GoogLeNet.tflite
GoogLeNet float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 1.564 ms 0 - 120 MB NPU GoogLeNet.dlc
GoogLeNet float SA7255P ADP Qualcomm® SA7255P TFLITE 5.022 ms 0 - 127 MB NPU GoogLeNet.tflite
GoogLeNet float SA7255P ADP Qualcomm® SA7255P QNN_DLC 5.058 ms 1 - 122 MB NPU GoogLeNet.dlc
GoogLeNet float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.841 ms 0 - 2 MB NPU GoogLeNet.tflite
GoogLeNet float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.835 ms 1 - 3 MB NPU GoogLeNet.dlc
GoogLeNet float SA8295P ADP Qualcomm® SA8295P TFLITE 1.815 ms 0 - 134 MB NPU GoogLeNet.tflite
GoogLeNet float SA8295P ADP Qualcomm® SA8295P QNN_DLC 1.797 ms 0 - 127 MB NPU GoogLeNet.dlc
GoogLeNet float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.845 ms 0 - 3 MB NPU GoogLeNet.tflite
GoogLeNet float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.84 ms 1 - 3 MB NPU GoogLeNet.dlc
GoogLeNet float SA8775P ADP Qualcomm® SA8775P TFLITE 1.584 ms 0 - 128 MB NPU GoogLeNet.tflite
GoogLeNet float SA8775P ADP Qualcomm® SA8775P QNN_DLC 1.564 ms 0 - 120 MB NPU GoogLeNet.dlc
GoogLeNet float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.58 ms 0 - 155 MB NPU GoogLeNet.tflite
GoogLeNet float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.583 ms 1 - 146 MB NPU GoogLeNet.dlc
GoogLeNet float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.678 ms 0 - 118 MB NPU GoogLeNet.onnx.zip
GoogLeNet float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.455 ms 0 - 132 MB NPU GoogLeNet.tflite
GoogLeNet float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.452 ms 0 - 124 MB NPU GoogLeNet.dlc
GoogLeNet float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.57 ms 0 - 99 MB NPU GoogLeNet.onnx.zip
GoogLeNet float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.379 ms 0 - 131 MB NPU GoogLeNet.tflite
GoogLeNet float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.382 ms 0 - 125 MB NPU GoogLeNet.dlc
GoogLeNet float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.507 ms 1 - 98 MB NPU GoogLeNet.onnx.zip
GoogLeNet float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.994 ms 1 - 1 MB NPU GoogLeNet.dlc
GoogLeNet float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.04 ms 13 - 13 MB NPU GoogLeNet.onnx.zip
GoogLeNet w8a8 Dragonwing Q-6690 MTP Qualcomm® Qcm6690 TFLITE 2.28 ms 0 - 128 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 Dragonwing Q-6690 MTP Qualcomm® Qcm6690 QNN_DLC 2.271 ms 0 - 129 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Dragonwing Q-6690 MTP Qualcomm® Qcm6690 ONNX 9.186 ms 8 - 22 MB CPU GoogLeNet.onnx.zip
GoogLeNet w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 TFLITE 0.969 ms 0 - 8 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 1.043 ms 2 - 4 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 13.109 ms 6 - 19 MB CPU GoogLeNet.onnx.zip
GoogLeNet w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 0.846 ms 0 - 121 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 0.805 ms 0 - 122 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 0.459 ms 0 - 142 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 0.42 ms 0 - 143 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.252 ms 0 - 4 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.243 ms 0 - 4 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 0.491 ms 0 - 9 MB NPU GoogLeNet.onnx.zip
GoogLeNet w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 0.438 ms 0 - 121 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 1.513 ms 0 - 122 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 13.018 ms 0 - 42 MB GPU GoogLeNet.tflite
GoogLeNet w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 7.791 ms 6 - 15 MB CPU GoogLeNet.onnx.zip
GoogLeNet w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 0.846 ms 0 - 121 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 0.805 ms 0 - 122 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.246 ms 0 - 3 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.252 ms 0 - 3 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 0.637 ms 0 - 128 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 0.614 ms 0 - 128 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.252 ms 0 - 3 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.249 ms 0 - 2 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 0.438 ms 0 - 121 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 1.513 ms 0 - 122 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.19 ms 0 - 145 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.182 ms 0 - 144 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.321 ms 0 - 124 MB NPU GoogLeNet.onnx.zip
GoogLeNet w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.153 ms 0 - 125 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.149 ms 0 - 126 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.286 ms 0 - 103 MB NPU GoogLeNet.onnx.zip
GoogLeNet w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile TFLITE 0.339 ms 0 - 128 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 0.335 ms 0 - 129 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 8.674 ms 8 - 26 MB CPU GoogLeNet.onnx.zip
GoogLeNet w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.142 ms 0 - 125 MB NPU GoogLeNet.tflite
GoogLeNet w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.137 ms 0 - 125 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.277 ms 0 - 103 MB NPU GoogLeNet.onnx.zip
GoogLeNet w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.356 ms 0 - 0 MB NPU GoogLeNet.dlc
GoogLeNet w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.4 ms 7 - 7 MB NPU GoogLeNet.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.googlenet.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.googlenet.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.googlenet.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.googlenet 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.googlenet.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.googlenet.demo -- --eval-mode 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 GoogLeNet's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of GoogLeNet can be found here.

References

Community

Downloads last month
185
Inference Providers NEW
Image Classification
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for qualcomm/GoogLeNet