Hugging Face's logo

qualcomm
/
Detectron2-Detection

Object Detection
PyTorch
LiteRT
android
Model card Files
xet
Community
Detectron2-Detection / README.md
qaihm-bot's picture
qaihm-bot
v0.45.0
e35bf52 verified
preview code
|
raw
history blame contribute delete
35.5 kB
metadata 
library_name: pytorch
license: other
tags:
  - android
pipeline_tag: object-detection

Detectron2-Detection: Optimized for Mobile Deployment

A next-generation library for object detection

Detectron2-Detection is a machine learning model that can detect objects (trained on COCO dataset).

This model is an implementation of Detectron2-Detection found here.

This repository provides scripts to run Detectron2-Detection 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: faster_rcnn_R_50_C4_1x
    • Input resolution: 800x800
    • Number of parameters: 33.8M
    • Model size (float): 129 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
Detectron2ProposalGenerator float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 187.314 ms 0 - 426 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 88.096 ms 7 - 469 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 56.493 ms 7 - 10 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 69.755 ms 1 - 428 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float SA7255P ADP Qualcomm® SA7255P QNN_DLC 187.314 ms 0 - 426 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float SA8295P ADP Qualcomm® SA8295P TFLITE 62.02 ms 11 - 413 MB NPU Detectron2-Detection.tflite
Detectron2ProposalGenerator float SA8295P ADP Qualcomm® SA8295P QNN_DLC 79.497 ms 0 - 393 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float SA8775P ADP Qualcomm® SA8775P QNN_DLC 69.755 ms 1 - 428 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 38.672 ms 7 - 526 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 23.367 ms 9 - 438 MB NPU Detectron2-Detection.tflite
Detectron2ProposalGenerator float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 28.699 ms 7 - 404 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile TFLITE 16.65 ms 10 - 403 MB NPU Detectron2-Detection.tflite
Detectron2ProposalGenerator float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile QNN_DLC 23.922 ms 7 - 413 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 56.926 ms 7 - 7 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 330.557 ms 4 - 1333 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 100.504 ms 10 - 415 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 64.389 ms 10 - 13 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 445.931 ms 1 - 1330 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float SA7255P ADP Qualcomm® SA7255P QNN_DLC 330.557 ms 4 - 1333 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float SA8295P ADP Qualcomm® SA8295P TFLITE 761.98 ms 0 - 1533 MB CPU Detectron2-Detection.tflite
Detectron2ROIHead float SA8295P ADP Qualcomm® SA8295P QNN_DLC 89.6 ms 0 - 358 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float SA8775P ADP Qualcomm® SA8775P QNN_DLC 445.931 ms 1 - 1330 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 48.169 ms 10 - 1932 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 2229.127 ms 0 - 1721 MB NPU Detectron2-Detection.tflite
Detectron2ROIHead float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 40.828 ms 10 - 1336 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile TFLITE 2001.288 ms 0 - 1886 MB NPU Detectron2-Detection.tflite
Detectron2ROIHead float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile QNN_DLC 26.899 ms 10 - 1414 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 64.294 ms 10 - 10 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Dragonwing Q-6690 MTP Qualcomm® QCM6690 QNN_DLC 355.555 ms 4 - 512 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 153.72 ms 4 - 14 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 2995.84 ms 189 - 197 MB CPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a16 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 113.986 ms 2 - 391 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 69.449 ms 4 - 515 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 64.245 ms 4 - 7 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 19.569 ms 15 - 19 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a16 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 64.402 ms 1 - 436 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 SA7255P ADP Qualcomm® SA7255P QNN_DLC 113.986 ms 2 - 391 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 SA8295P ADP Qualcomm® SA8295P QNN_DLC 74.127 ms 0 - 434 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 SA8775P ADP Qualcomm® SA8775P QNN_DLC 64.402 ms 1 - 436 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 46.522 ms 4 - 492 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 13.528 ms 17 - 409 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 26.781 ms 4 - 391 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 10.64 ms 17 - 310 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a16 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 54.328 ms 4 - 445 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 1394.375 ms 173 - 189 MB CPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile QNN_DLC 22.864 ms 4 - 419 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile ONNX 8.655 ms 20 - 385 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a16 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 66.822 ms 4 - 4 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a16 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 20.186 ms 33 - 33 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 Dragonwing Q-6690 MTP Qualcomm® QCM6690 QNN_DLC 497.166 ms 5 - 473 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 130.402 ms 1 - 7 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 8643.155 ms 3039 - 3054 MB CPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 101.309 ms 4 - 364 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 57.782 ms 5 - 428 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 35.18 ms 5 - 7 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 303.607 ms 1014 - 1016 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 35.311 ms 0 - 361 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 SA7255P ADP Qualcomm® SA7255P QNN_DLC 101.309 ms 4 - 364 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 SA8295P ADP Qualcomm® SA8295P QNN_DLC 49.802 ms 0 - 362 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 SA8775P ADP Qualcomm® SA8775P QNN_DLC 35.311 ms 0 - 361 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 25.1 ms 5 - 427 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 262.813 ms 1008 - 1435 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 18.557 ms 5 - 391 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 239.171 ms 1001 - 1382 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 47.939 ms 5 - 420 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 3592.796 ms 3033 - 3048 MB CPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile QNN_DLC 14.274 ms 5 - 393 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile ONNX 188.861 ms 1014 - 1399 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a16 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 36.239 ms 5 - 5 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a16 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 256.69 ms 1059 - 1059 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 Dragonwing Q-6690 MTP Qualcomm® QCM6690 QNN_DLC 186.888 ms 2 - 377 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Dragonwing Q-6690 MTP Qualcomm® QCM6690 ONNX 449.325 ms 121 - 136 MB CPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 63.304 ms 2 - 8 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 447.641 ms 141 - 175 MB CPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 72.247 ms 2 - 371 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 47.461 ms 2 - 426 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 40.044 ms 2 - 4 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 12.406 ms 8 - 10 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 183.946 ms 1 - 371 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 72.247 ms 2 - 371 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 45.448 ms 2 - 375 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 183.946 ms 1 - 371 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 26.892 ms 2 - 436 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 8.924 ms 10 - 327 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 22.806 ms 2 - 358 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 7.066 ms 8 - 256 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 29.502 ms 2 - 379 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 357.966 ms 133 - 148 MB CPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile QNN_DLC 23.106 ms 2 - 367 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile ONNX 5.733 ms 8 - 258 MB NPU Detectron2-Detection.onnx.zip
Detectron2ProposalGenerator w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 42.189 ms 2 - 2 MB NPU Detectron2-Detection.dlc
Detectron2ProposalGenerator w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 12.51 ms 20 - 20 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 Dragonwing Q-6690 MTP Qualcomm® QCM6690 QNN_DLC 278.301 ms 3 - 372 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Dragonwing Q-6690 MTP Qualcomm® QCM6690 ONNX 1072.446 ms 2473 - 2487 MB CPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 80.145 ms 4 - 8 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 1721.984 ms 2531 - 2545 MB CPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 53.145 ms 2 - 300 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 40.959 ms 3 - 347 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 19.269 ms 2 - 5 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 112.829 ms 507 - 510 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 19.861 ms 2 - 302 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 53.145 ms 2 - 300 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 29.261 ms 0 - 297 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 19.861 ms 2 - 302 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 13.498 ms 2 - 354 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 93.183 ms 507 - 843 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 10.295 ms 2 - 310 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 88.013 ms 467 - 757 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 29.544 ms 2 - 324 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 983.931 ms 2509 - 2525 MB CPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile QNN_DLC 8.368 ms 2 - 310 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen 5 Mobile ONNX 68.356 ms 509 - 802 MB NPU Detectron2-Detection.onnx.zip
Detectron2ROIHead w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 19.97 ms 2 - 2 MB NPU Detectron2-Detection.dlc
Detectron2ROIHead w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 100.047 ms 530 - 530 MB NPU Detectron2-Detection.onnx.zip

Installation

Install the package via pip:

# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install wheel==0.45.1 "torch>=2.1,<2.9.0" "setuptools>=77.0.3"
pip install "qai-hub-models[detectron2-detection]" git+https://github.com/facebookresearch/detectron2.git@d38d716 --no-build-isolation

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.detectron2_detection.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.detectron2_detection.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.detectron2_detection.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.detectron2_detection 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.detectron2_detection.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.detectron2_detection.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 Detectron2-Detection's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Detectron2-Detection can be found here.

References

Community