File size: 6,191 Bytes
2941c2f 78602eb 2941c2f 78602eb 2941c2f 78602eb 2941c2f 78602eb 2941c2f 34b43b9 2941c2f 34b43b9 2941c2f | 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 | ---
library_name: pytorch
license: apache-2.0
tags:
- bu_auto
- android
pipeline_tag: other
---
# BEVFormer: Optimized for Qualcomm Devices
Bevformer is a SOTA model of interest to the Auto BU.
This is based on the implementation of BEVFormer found [here](https://github.com/fundamentalvision/BEVFormer/).
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/bevformer) 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/bevformer/releases/v0.66.0/bevformer-onnx-float.zip)
| QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/bevformer/releases/v0.66.0/bevformer-qnn_dlc-float.zip)
For more device-specific assets and performance metrics, visit **[BEVFormer on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/bevformer)**.
### 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/bevformer) 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 [BEVFormer on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.66.0/src/qai_hub_models/models/bevformer) for usage instructions.
## Model Details
**Model Type:** Model_use_case.driver_assistance
**Model Stats:**
- Input preprocessing: Images must be ImageNet mean/std normalized before inference; the model does not normalize internally.
- Input resolution: 6 x 3 x 480 x 800
- Model checkpoint: bevformer_tiny_deformable_optimized_exp_86_epoch_24.pth
- Model size: 120 MB
- Number of parameters: 27M
## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| BEVFormer | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 898.972 ms | 7 - 786 MB | NPU
| BEVFormer | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1111.845 ms | 3 - 739 MB | NPU
| BEVFormer | ONNX | float | Snapdragon® X2 Elite | 887.121 ms | 29 - 29 MB | NPU
| BEVFormer | ONNX | float | Snapdragon® X Elite | 1707.571 ms | 50 - 50 MB | NPU
| BEVFormer | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1354.737 ms | 0 - 972 MB | NPU
| BEVFormer | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 1769.929 ms | 29 - 986 MB | NPU
| BEVFormer | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 1920.481 ms | 29 - 62 MB | NPU
| BEVFormer | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1642.585 ms | 0 - 52 MB | NPU
| BEVFormer | ONNX | float | Qualcomm® QCS8450 | 1769.929 ms | 29 - 986 MB | NPU
| BEVFormer | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 1827.543 ms | 29 - 61 MB | NPU
| BEVFormer | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 1707.571 ms | 50 - 50 MB | NPU
| BEVFormer | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 1111.845 ms | 3 - 739 MB | NPU
| BEVFormer | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 880.995 ms | 20 - 860 MB | NPU
| BEVFormer | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1438.781 ms | 10 - 801 MB | NPU
| BEVFormer | QNN_DLC | float | Snapdragon® X2 Elite | 924.225 ms | 29 - 29 MB | NPU
| BEVFormer | QNN_DLC | float | Snapdragon® X Elite | 1703.876 ms | 29 - 29 MB | NPU
| BEVFormer | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1342.843 ms | 27 - 1002 MB | NPU
| BEVFormer | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 1415.097 ms | 28 - 1013 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 1996.15 ms | 29 - 63 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1584.595 ms | 29 - 32 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® SA8650P | 2011.958 ms | 26 - 863 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® SA8255P | 2011.958 ms | 26 - 863 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® QCS8450 | 1415.097 ms | 28 - 1013 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 1789.621 ms | 29 - 62 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 1703.876 ms | 29 - 29 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1438.781 ms | 10 - 801 MB | NPU
| BEVFormer | QNN_DLC | float | Qualcomm® SA8295P | 1739.194 ms | 26 - 720 MB | NPU
## License
* The license for the original implementation of BEVFormer can be found
[here](https://github.com/fundamentalvision/BEVFormer/blob/master/LICENSE).
## References
* [BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers](https://arxiv.org/abs/2203.17270)
* [Source Model Implementation](https://github.com/fundamentalvision/BEVFormer/)
## 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).
|