File size: 12,301 Bytes
29d6673 9d7668d 29d6673 914e545 29d6673 c5ae9b8 29d6673 1a33e82 29d6673 7d99418 b86b199 29d6673 7d99418 74cf852 7d99418 74cf852 7d99418 74cf852 7d99418 74cf852 7d99418 9d58c9b 7d99418 9d58c9b 7d99418 9fef77b 8515975 29d6673 f0061b4 8515975 29d6673 cfb899a 29d6673 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 | ---
library_name: pytorch
license: bsd-3-clause
tags:
- backbone
- bu_auto
- android
pipeline_tag: image-classification
---

# ResNet50: Optimized for Qualcomm Devices
ResNet50 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
This is based on the implementation of ResNet50 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py).
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/v0.66.0/src/qai_hub_models/models/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 2.50, ONNX Runtime 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.66.0/resnet50-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.50, ONNX Runtime 1.30.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.66.0/resnet50-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.66.0/resnet50-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.66.0/resnet50-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.66.0/resnet50-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet50/releases/v0.66.0/resnet50-tflite-w8a8.zip)
For more device-specific assets and performance metrics, visit **[ResNet50 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/resnet50)**.
### Option 2: Export with Custom Configurations
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.66.0/src/qai_hub_models/models/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 [ResNet50 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.66.0/src/qai_hub_models/models/resnet50) for usage instructions.
## Model Details
**Model Type:** Model_use_case.image_classification
**Model Stats:**
- Input resolution: 224x224
- Model checkpoint: Imagenet
- Model size (float): 97.4 MB
- Model size (w8a8): 25.1 MB
- Number of parameters: 25.5M
## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ResNet50 | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.07 ms | 0 - 53 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.224 ms | 0 - 50 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® X2 Elite | 0.964 ms | 2 - 2 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® X Elite | 1.922 ms | 50 - 50 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.434 ms | 0 - 82 MB | NPU
| ResNet50 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 2.889 ms | 0 - 67 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 3.3 ms | 0 - 5 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.981 ms | 0 - 71 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® QCS8450 | 2.889 ms | 0 - 67 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 3.0 ms | 1 - 4 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.922 ms | 50 - 50 MB | NPU
| ResNet50 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1.224 ms | 0 - 50 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.559 ms | 0 - 58 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.578 ms | 0 - 53 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® X2 Elite | 0.409 ms | 1 - 1 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® X Elite | 0.843 ms | 25 - 25 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.674 ms | 0 - 89 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.055 ms | 0 - 90 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 2.64 ms | 0 - 3 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.977 ms | 0 - 4 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.91 ms | 0 - 30 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® QCS8450 | 1.055 ms | 0 - 90 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.018 ms | 0 - 3 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.843 ms | 25 - 25 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 6.219 ms | 0 - 170 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.134 ms | 0 - 59 MB | NPU
| ResNet50 | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.578 ms | 0 - 53 MB | NPU
| ResNet50 | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.134 ms | 0 - 59 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.096 ms | 1 - 53 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.281 ms | 0 - 47 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® X2 Elite | 1.225 ms | 1 - 1 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® X Elite | 2.348 ms | 1 - 1 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.617 ms | 0 - 79 MB | NPU
| ResNet50 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 3.475 ms | 0 - 67 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 3.36 ms | 1 - 4 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.245 ms | 1 - 2 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8775P | 3.316 ms | 1 - 50 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8650P | 3.316 ms | 1 - 50 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8255P | 3.316 ms | 1 - 50 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® QCS8450 | 3.475 ms | 0 - 67 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 3.168 ms | 3 - 5 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.348 ms | 1 - 1 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.281 ms | 0 - 47 MB | NPU
| ResNet50 | QNN_DLC | float | Qualcomm® SA8295P | 3.548 ms | 0 - 32 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.486 ms | 0 - 48 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.534 ms | 0 - 51 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.444 ms | 0 - 0 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.939 ms | 0 - 0 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.657 ms | 0 - 77 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.054 ms | 0 - 82 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 2.96 ms | 2 - 4 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.943 ms | 0 - 3 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.901 ms | 0 - 2 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.086 ms | 0 - 52 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.086 ms | 0 - 52 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.086 ms | 0 - 52 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.054 ms | 0 - 82 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.965 ms | 0 - 2 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.939 ms | 0 - 0 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 6.443 ms | 0 - 161 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.222 ms | 0 - 55 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.534 ms | 0 - 51 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Qualcomm® SA8295P | 1.309 ms | 0 - 47 MB | NPU
| ResNet50 | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.222 ms | 0 - 55 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 1.107 ms | 0 - 72 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.257 ms | 0 - 69 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.592 ms | 0 - 118 MB | NPU
| ResNet50 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 3.471 ms | 0 - 103 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 3.334 ms | 0 - 53 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.203 ms | 0 - 3 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® QCS8450 | 3.471 ms | 0 - 103 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 3.132 ms | 0 - 52 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.257 ms | 0 - 69 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® SA7255P | 10.595 ms | 0 - 71 MB | NPU
| ResNet50 | TFLITE | float | Qualcomm® SA8295P | 3.501 ms | 0 - 54 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 0.453 ms | 0 - 46 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.477 ms | 0 - 43 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.585 ms | 0 - 78 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.921 ms | 0 - 86 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 2.662 ms | 0 - 27 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 0.835 ms | 0 - 28 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.752 ms | 0 - 2 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® QCS8450 | 0.921 ms | 0 - 86 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.839 ms | 0 - 27 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 5.867 ms | 0 - 157 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.08 ms | 0 - 53 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.477 ms | 0 - 43 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® SA7255P | 1.759 ms | 0 - 47 MB | NPU
| ResNet50 | TFLITE | w8a8 | Qualcomm® SA8295P | 1.154 ms | 0 - 45 MB | NPU
| ResNet50 | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.08 ms | 0 - 53 MB | NPU
## License
* The license for the original implementation of ResNet50 can be found
[here](https://github.com/pytorch/vision/blob/main/LICENSE).
## References
* [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py)
## 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:[email protected]).
|