| --- |
| library_name: pytorch |
| license: other |
| tags: |
| - bu_auto |
| - android |
| pipeline_tag: image-classification |
|
|
| --- |
| |
|  |
|
|
| # NASNet: Optimized for Qualcomm Devices |
|
|
| NASNet is a CNN-based architecture discovered via Neural Architecture Search (NAS) that can classify images from the Imagenet dataset. |
|
|
| This is based on the implementation of NASNet found [here](https://github.com/huggingface/pytorch-image-models/tree/main). |
| 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/nasnet) 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/nasnet/releases/v0.51.0/nasnet-onnx-float.zip) |
| | ONNX | w8a8_mixed_fp16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.51.0/nasnet-onnx-w8a8_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/nasnet/releases/v0.51.0/nasnet-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/nasnet/releases/v0.51.0/nasnet-tflite-float.zip) |
| |
| For more device-specific assets and performance metrics, visit **[NASNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/nasnet)**. |
| |
| |
| ### 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/nasnet) 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 [NASNet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/nasnet) for usage instructions. |
| |
| ## Model Details |
| |
| **Model Type:** Model_use_case.image_classification |
|
|
| **Model Stats:** |
| - Model checkpoint: nasnetalarge.tf_in1k |
| - Input resolution: 224x224 |
| - GMACs: 5.9 |
| - Activations (M): 19.4 |
| - Number of parameters: 88.7M |
| - Model size (float): 338 MB |
| |
| ## Performance Summary |
| | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
| |---|---|---|---|---|---|--- |
| | NASNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.372 ms | 1 - 677 MB | NPU |
| | NASNet | ONNX | float | Snapdragon® X2 Elite | 8.998 ms | 189 - 189 MB | NPU |
| | NASNet | ONNX | float | Snapdragon® X Elite | 17.761 ms | 188 - 188 MB | NPU |
| | NASNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 12.727 ms | 1 - 834 MB | NPU |
| | NASNet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 17.503 ms | 0 - 194 MB | NPU |
| | NASNet | ONNX | float | Qualcomm® QCS9075 | 27.992 ms | 0 - 4 MB | NPU |
| | NASNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 10.678 ms | 0 - 637 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® 8 Elite Gen 5 Mobile | 5.525 ms | 11 - 389 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® X2 Elite | 5.675 ms | 100 - 100 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® X Elite | 13.241 ms | 98 - 98 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® 8 Gen 3 Mobile | 7.977 ms | 12 - 515 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Qualcomm® QCS8550 (Proxy) | 11.416 ms | 9 - 485 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Qualcomm® QCS9075 | 13.77 ms | 11 - 14 MB | NPU |
| | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® 8 Elite For Galaxy Mobile | 6.824 ms | 12 - 391 MB | NPU |
| | NASNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 9.064 ms | 1 - 662 MB | NPU |
| | NASNet | QNN_DLC | float | Snapdragon® X2 Elite | 9.429 ms | 1 - 1 MB | NPU |
| | NASNet | QNN_DLC | float | Snapdragon® X Elite | 17.926 ms | 1 - 1 MB | NPU |
| | NASNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 12.289 ms | 0 - 834 MB | NPU |
| | NASNet | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 53.776 ms | 1 - 666 MB | NPU |
| | NASNet | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 17.107 ms | 1 - 3 MB | NPU |
| | NASNet | QNN_DLC | float | Qualcomm® QCS9075 | 29.112 ms | 1 - 3 MB | NPU |
| | NASNet | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 35.015 ms | 0 - 808 MB | NPU |
| | NASNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 10.42 ms | 1 - 706 MB | NPU |
| | NASNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.552 ms | 0 - 575 MB | NPU |
| | NASNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 8.676 ms | 0 - 711 MB | NPU |
| | NASNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 43.227 ms | 0 - 578 MB | NPU |
| | NASNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 12.226 ms | 0 - 3 MB | NPU |
| | NASNet | TFLITE | float | Qualcomm® QCS9075 | 15.146 ms | 0 - 191 MB | NPU |
| | NASNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 28.859 ms | 0 - 695 MB | NPU |
| | NASNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.922 ms | 0 - 562 MB | NPU |
|
|
| ## License |
| * The license for the original implementation of NASNet can be found |
| [here](https://github.com/huggingface/pytorch-image-models?tab=Apache-2.0-1-ov-file). |
|
|
| ## References |
| * [Learning Transferable Architectures for Scalable Image Recognition](https://arxiv.org/abs/1707.07012) |
| * [Source Model Implementation](https://github.com/huggingface/pytorch-image-models/tree/main) |
|
|
| ## 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). |
|
|
