README.md

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

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/conditional_detr_resnet50/web-assets/model_demo.png)

# Conditional-DETR-ResNet50: 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 Conditional-DETR-ResNet50 found [here](https://github.com/huggingface/transformers/tree/main/src/transformers/models/conditional_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/v0.59.0/src/qai_hub_models/models/conditional_detr_resnet50) 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 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/conditional_detr_resnet50/releases/v0.59.0/conditional_detr_resnet50-onnx-float.zip)
| ONNX | w8a16_mixed_fp16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/conditional_detr_resnet50/releases/v0.59.0/conditional_detr_resnet50-onnx-w8a16_mixed_fp16.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/conditional_detr_resnet50/releases/v0.59.0/conditional_detr_resnet50-qnn_dlc-float.zip)
| QNN_DLC | w8a16_mixed_fp16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/conditional_detr_resnet50/releases/v0.59.0/conditional_detr_resnet50-qnn_dlc-w8a16_mixed_fp16.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/conditional_detr_resnet50/releases/v0.59.0/conditional_detr_resnet50-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[Conditional-DETR-ResNet50 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/conditional_detr_resnet50)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/conditional_detr_resnet50) 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 [Conditional-DETR-ResNet50 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/conditional_detr_resnet50) for usage instructions.

## Model Details

**Model Type:** Model_use_case.object_detection

**Model Stats:**
- Model checkpoint: ResNet50
- Input resolution: 480x480
- Number of parameters: 43.6M
- Model size (float): 166 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| Conditional-DETR-ResNet50 | ONNX | float | Snapdragon® X2 Elite | 9.295 ms | 5 - 5 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Snapdragon® X Elite | 20.416 ms | 83 - 83 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 15.028 ms | 1 - 455 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 39.424 ms | 1 - 368 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 20.145 ms | 0 - 524 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Qualcomm® QCS8450 | 39.424 ms | 1 - 368 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 29.705 ms | 5 - 12 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 20.416 ms | 83 - 83 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 11.249 ms | 0 - 345 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Snapdragon® 8 Elite Mobile | 11.249 ms | 0 - 345 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.288 ms | 3 - 321 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | w8a16_mixed_fp16 | Snapdragon® 8 Gen 3 Mobile | 65.69 ms | 17 - 492 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | w8a16_mixed_fp16 | Qualcomm® Dragonwing™ Q-8750 | 53.631 ms | 23 - 398 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | w8a16_mixed_fp16 | Snapdragon® 8 Elite Mobile | 53.631 ms | 23 - 398 MB | NPU
| Conditional-DETR-ResNet50 | ONNX | w8a16_mixed_fp16 | Snapdragon® 8 Elite Gen 5 Mobile | 43.242 ms | 22 - 438 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Snapdragon® X2 Elite | 10.054 ms | 5 - 5 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Snapdragon® X Elite | 23.399 ms | 5 - 5 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 17.021 ms | 0 - 397 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 45.071 ms | 5 - 351 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 97.924 ms | 0 - 309 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 23.155 ms | 5 - 391 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® SA8775P | 32.102 ms | 0 - 308 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® SA8650P | 32.102 ms | 0 - 308 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® SA8255P | 32.102 ms | 0 - 308 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® QCS8450 | 45.071 ms | 5 - 351 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 32.372 ms | 7 - 13 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 23.399 ms | 5 - 5 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 11.592 ms | 1 - 317 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® SA7255P | 97.924 ms | 0 - 309 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Qualcomm® SA8295P | 34.07 ms | 2 - 263 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 11.592 ms | 1 - 317 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.967 ms | 5 - 338 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Snapdragon® X2 Elite | 18.833 ms | 2 - 2 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Snapdragon® X Elite | 37.486 ms | 2 - 2 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Snapdragon® 8 Gen 3 Mobile | 26.603 ms | 2 - 504 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® Dragonwing™ QCS8275 | 105.201 ms | 2 - 401 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 37.402 ms | 2 - 567 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® SA8775P | 43.315 ms | 1 - 400 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® SA8650P | 43.315 ms | 1 - 400 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® SA8255P | 43.315 ms | 1 - 400 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® Dragonwing™ IQ-9075 | 43.446 ms | 2 - 6 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® Dragonwing™ IQ-X7181 | 37.486 ms | 2 - 2 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® Dragonwing™ Q-8750 | 20.158 ms | 0 - 407 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Qualcomm® SA7255P | 105.201 ms | 2 - 401 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Snapdragon® 8 Elite Mobile | 20.158 ms | 0 - 407 MB | NPU
| Conditional-DETR-ResNet50 | QNN_DLC | w8a16_mixed_fp16 | Snapdragon® 8 Elite Gen 5 Mobile | 17.236 ms | 1 - 453 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 15.646 ms | 0 - 435 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 44.806 ms | 0 - 374 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 93.301 ms | 0 - 337 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 22.055 ms | 0 - 3 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® SA8775P | 29.971 ms | 0 - 368 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® SA8650P | 29.971 ms | 0 - 368 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® SA8255P | 29.971 ms | 0 - 368 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® QCS8450 | 44.806 ms | 0 - 374 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 30.193 ms | 0 - 93 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 11.903 ms | 0 - 340 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® SA7255P | 93.301 ms | 0 - 337 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Qualcomm® SA8295P | 34.67 ms | 0 - 289 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Snapdragon® 8 Elite Mobile | 11.903 ms | 0 - 340 MB | NPU
| Conditional-DETR-ResNet50 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 9.309 ms | 0 - 356 MB | NPU

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

## References
* [Conditional {DETR} for Fast Training Convergence](https://arxiv.org/abs/2108.06152)
* [Source Model Implementation](https://github.com/huggingface/transformers/tree/main/src/transformers/models/conditional_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).