---
library_name: pytorch
license: other
tags:
- bu_auto
- android
pipeline_tag: object-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detr_resnet50_dc5/web-assets/model_demo.png)

# DETR-ResNet50-DC5: Optimized for Qualcomm Devices

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

This is based on the implementation of DETR-ResNet50-DC5 found [here](https://github.com/facebookresearch/detr).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/detr_resnet50_dc5) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detr_resnet50_dc5/releases/v0.51.0/detr_resnet50_dc5-onnx-float.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/detr_resnet50_dc5/releases/v0.51.0/detr_resnet50_dc5-qnn_dlc-float.zip)

For more device-specific assets and performance metrics, visit **[DETR-ResNet50-DC5 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/detr_resnet50_dc5)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/detr_resnet50_dc5) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [DETR-ResNet50-DC5 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/detr_resnet50_dc5) for usage instructions.

## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- Model checkpoint: ResNet50-DC5
- Input resolution: 480x480
- Model size (float): 160 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| DETR-ResNet50-DC5 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 20.495 ms | 5 - 497 MB | NPU
| DETR-ResNet50-DC5 | ONNX | float | Snapdragon® X2 Elite | 21.931 ms | 79 - 79 MB | NPU
| DETR-ResNet50-DC5 | ONNX | float | Snapdragon® X Elite | 47.249 ms | 78 - 78 MB | NPU
| DETR-ResNet50-DC5 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 32.059 ms | 1 - 647 MB | NPU
| DETR-ResNet50-DC5 | ONNX | float | Qualcomm® QCS8550 (Proxy) | 45.368 ms | 5 - 9 MB | NPU
| DETR-ResNet50-DC5 | ONNX | float | Qualcomm® QCS9075 | 68.446 ms | 5 - 12 MB | NPU
| DETR-ResNet50-DC5 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 23.794 ms | 3 - 463 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 24.84 ms | 5 - 532 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Snapdragon® X2 Elite | 25.17 ms | 5 - 5 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Snapdragon® X Elite | 55.355 ms | 5 - 5 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 39.914 ms | 5 - 677 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 185.999 ms | 1 - 499 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 54.12 ms | 5 - 7 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® SA8775P | 67.753 ms | 1 - 501 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® QCS9075 | 80.529 ms | 7 - 13 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 95.527 ms | 4 - 528 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® SA7255P | 185.999 ms | 1 - 499 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Qualcomm® SA8295P | 82.112 ms | 0 - 406 MB | NPU
| DETR-ResNet50-DC5 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 27.527 ms | 0 - 520 MB | NPU

## License
* The license for the original implementation of DETR-ResNet50-DC5 can be found
  [here](https://github.com/facebookresearch/detr/blob/main/LICENSE).

## References
* [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872)
* [Source Model Implementation](https://github.com/facebookresearch/detr)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).