File size: 6,601 Bytes
db6062e 15a8ad4 db6062e b6e68ca db6062e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 | ---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: object-detection
---
# ResNet34-SSD: Optimized for Qualcomm Devices
ResNet34-SSD is a single-stage object detection model that integrates the ResNet34 backbone with the SSD (Single Shot MultiBox Detector) framework. It is optimized for real-time detection tasks and supports multiple deployment backends including PyTorch, TensorFlow, and ONNX.
This is based on the implementation of ResNet34-SSD found [here](https://github.com/mlcommons/inference/tree/33894a19c4af6207f7cfdda75f84570f04836de5/vision/classification_and_detection).
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/resnet34_ssd1200) 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/resnet34_ssd1200/releases/v0.53.1/resnet34_ssd1200-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/resnet34_ssd1200/releases/v0.53.1/resnet34_ssd1200-qnn_dlc-float.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet34_ssd1200/releases/v0.53.1/resnet34_ssd1200-tflite-float.zip)
For more device-specific assets and performance metrics, visit **[ResNet34-SSD on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/resnet34_ssd1200)**.
### 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/resnet34_ssd1200) 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 [ResNet34-SSD on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/resnet34_ssd1200) for usage instructions.
## Model Details
**Model Type:** Model_use_case.object_detection
**Model Stats:**
- Model checkpoint: resnet34-ssd1200
- Input resolution: 1x3x1200x1200
- Number of parameters: 20.0M
- Model size (float): 76.2 MB
## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ResNet34-SSD | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 38.733 ms | 17 - 513 MB | NPU
| ResNet34-SSD | ONNX | float | Snapdragon® 8 Elite Mobile | 50.177 ms | 2 - 428 MB | NPU
| ResNet34-SSD | ONNX | float | Snapdragon® X2 Elite | 43.419 ms | 30 - 30 MB | NPU
| ResNet34-SSD | ONNX | float | Snapdragon® X Elite | 91.464 ms | 29 - 29 MB | NPU
| ResNet34-SSD | ONNX | float | Snapdragon® X Elite | 91.464 ms | 29 - 29 MB | NPU
| ResNet34-SSD | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 62.899 ms | 0 - 517 MB | NPU
| ResNet34-SSD | ONNX | float | Qualcomm® QCS8550 (Proxy) | 90.698 ms | 0 - 573 MB | NPU
| ResNet34-SSD | ONNX | float | Qualcomm® QCS9075 | 152.711 ms | 16 - 36 MB | NPU
| ResNet34-SSD | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 50.177 ms | 2 - 428 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 52.76 ms | 15 - 553 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 66.869 ms | 16 - 391 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® X2 Elite | 61.836 ms | 17 - 17 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® X Elite | 128.96 ms | 17 - 17 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® X Elite | 128.96 ms | 17 - 17 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 84.568 ms | 15 - 606 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 481.914 ms | 16 - 385 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 128.396 ms | 17 - 19 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Qualcomm® QCS9075 | 193.951 ms | 17 - 35 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 262.739 ms | 3 - 511 MB | NPU
| ResNet34-SSD | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 66.869 ms | 16 - 391 MB | NPU
| ResNet34-SSD | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 76.117 ms | 0 - 564 MB | NPU
| ResNet34-SSD | TFLITE | float | Snapdragon® 8 Elite Mobile | 88.158 ms | 0 - 402 MB | NPU
| ResNet34-SSD | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 107.7 ms | 0 - 543 MB | NPU
| ResNet34-SSD | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 513.324 ms | 0 - 378 MB | NPU
| ResNet34-SSD | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 146.036 ms | 0 - 3 MB | NPU
| ResNet34-SSD | TFLITE | float | Qualcomm® QCS9075 | 199.375 ms | 0 - 64 MB | NPU
| ResNet34-SSD | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 234.949 ms | 1 - 617 MB | NPU
| ResNet34-SSD | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 88.158 ms | 0 - 402 MB | NPU
## License
* The license for the original implementation of ResNet34-SSD can be found
[here](https://github.com/mlcommons/inference/blob/33894a19c4af6207f7cfdda75f84570f04836de5/LICENSE.md).
## References
* [Source Model Implementation](https://github.com/mlcommons/inference/tree/33894a19c4af6207f7cfdda75f84570f04836de5/vision/classification_and_detection)
## 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).
|