Hugging Face's logo

qualcomm
/
DETR-ResNet101

Object Detection
PyTorch
LiteRT
ONNX
android
Model card Files Community

DETR-ResNet101: Optimized for Mobile Deployment

Transformer based object detector with ResNet101 backbone

DETR is a machine learning model that can detect objects (trained on COCO dataset).

This model is an implementation of DETR-ResNet101 found here.

This repository provides scripts to run DETR-ResNet101 on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.object_detection
  • Model Stats:
    • Model checkpoint: ResNet101
    • Input resolution: 480x480
    • Number of parameters: 60.3M
    • Model size: 230 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
DETR-ResNet101 float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 139.96 ms 0 - 175 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN 137.552 ms 1 - 10 MB NPU Use Export Script
DETR-ResNet101 float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 39.3 ms 0 - 165 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN 48.852 ms 5 - 98 MB NPU Use Export Script
DETR-ResNet101 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 26.497 ms 0 - 31 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN 26.838 ms 5 - 8 MB NPU Use Export Script
DETR-ResNet101 float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 41.874 ms 0 - 175 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN 40.599 ms 1 - 11 MB NPU Use Export Script
DETR-ResNet101 float SA7255P ADP Qualcomm® SA7255P TFLITE 139.96 ms 0 - 175 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float SA7255P ADP Qualcomm® SA7255P QNN 137.552 ms 1 - 10 MB NPU Use Export Script
DETR-ResNet101 float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 26.523 ms 0 - 27 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN 26.571 ms 7 - 9 MB NPU Use Export Script
DETR-ResNet101 float SA8295P ADP Qualcomm® SA8295P TFLITE 45.449 ms 0 - 140 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float SA8295P ADP Qualcomm® SA8295P QNN 45.112 ms 0 - 18 MB NPU Use Export Script
DETR-ResNet101 float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 26.655 ms 0 - 29 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN 26.514 ms 5 - 7 MB NPU Use Export Script
DETR-ResNet101 float SA8775P ADP Qualcomm® SA8775P TFLITE 41.874 ms 0 - 175 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float SA8775P ADP Qualcomm® SA8775P QNN 40.599 ms 1 - 11 MB NPU Use Export Script
DETR-ResNet101 float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 26.668 ms 0 - 33 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN 26.588 ms 5 - 51 MB NPU Use Export Script
DETR-ResNet101 float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 25.034 ms 1 - 334 MB NPU DETR-ResNet101.onnx
DETR-ResNet101 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 18.939 ms 355 - 552 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN 19.492 ms 5 - 139 MB NPU Use Export Script
DETR-ResNet101 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 18.491 ms 127 - 281 MB NPU DETR-ResNet101.onnx
DETR-ResNet101 float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 17.962 ms 0 - 174 MB NPU DETR-ResNet101.tflite
DETR-ResNet101 float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN 16.859 ms 5 - 135 MB NPU Use Export Script
DETR-ResNet101 float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 14.893 ms 5 - 132 MB NPU DETR-ResNet101.onnx
DETR-ResNet101 float Snapdragon X Elite CRD Snapdragon® X Elite QNN 26.761 ms 5 - 5 MB NPU Use Export Script
DETR-ResNet101 float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 26.491 ms 115 - 115 MB NPU DETR-ResNet101.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[detr-resnet101]"

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.detr_resnet101.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.detr_resnet101.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.detr_resnet101.export

Profiling Results
------------------------------------------------------------
DETR-ResNet101
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 140.0                                
Estimated peak memory usage (MB): [0, 175]                             
Total # Ops                     : 874                                  
Compute Unit(s)                 : npu (874 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.detr_resnet101 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.detr_resnet101.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.detr_resnet101.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 DETR-ResNet101's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of DETR-ResNet101 can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month
102
Inference Providers NEW
Object Detection
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support