--- library_name: pytorch license: other tags: - android pipeline_tag: depth-estimation --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v2/web-assets/model_demo.png) # Depth-Anything-V2: Optimized for Qualcomm Devices Depth Anything is designed for estimating depth at each point in an image. This is based on the implementation of Depth-Anything-V2 found [here](https://github.com/huggingface/transformers/tree/main/src/transformers/models/depth_anything). 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/depth_anything_v2) 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/depth_anything_v2/releases/v0.51.0/depth_anything_v2-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/depth_anything_v2/releases/v0.51.0/depth_anything_v2-qnn_dlc-float.zip) | QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v2/releases/v0.51.0/depth_anything_v2-qnn_dlc-w8a16.zip) For more device-specific assets and performance metrics, visit **[Depth-Anything-V2 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/depth_anything_v2)**. ### 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/depth_anything_v2) 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 [Depth-Anything-V2 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/depth_anything_v2) for usage instructions. ## Model Details **Model Type:** Model_use_case.depth_estimation **Model Stats:** - Model checkpoint: DepthAnything_V2_Small - Input resolution: 518x518 - Number of parameters: 24.7M - Model size (float): 94.3 MB ## Performance Summary | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |---|---|---|---|---|---|--- | Depth-Anything-V2 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 35.293 ms | 5 - 704 MB | NPU | Depth-Anything-V2 | ONNX | float | Snapdragon® X2 Elite | 34.367 ms | 52 - 52 MB | NPU | Depth-Anything-V2 | ONNX | float | Snapdragon® X Elite | 75.623 ms | 51 - 51 MB | NPU | Depth-Anything-V2 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 52.9 ms | 0 - 937 MB | NPU | Depth-Anything-V2 | ONNX | float | Qualcomm® QCS8550 (Proxy) | 80.376 ms | 0 - 55 MB | NPU | Depth-Anything-V2 | ONNX | float | Qualcomm® QCS9075 | 90.888 ms | 3 - 9 MB | NPU | Depth-Anything-V2 | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 45.732 ms | 2 - 785 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 37.589 ms | 3 - 518 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Snapdragon® X2 Elite | 38.653 ms | 3 - 3 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Snapdragon® X Elite | 94.036 ms | 3 - 3 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 73.253 ms | 3 - 697 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 207.978 ms | 0 - 515 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 96.771 ms | 3 - 5 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® SA8775P | 100.748 ms | 1 - 507 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® QCS9075 | 101.521 ms | 3 - 9 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 174.105 ms | 1 - 586 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® SA7255P | 207.978 ms | 0 - 515 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Qualcomm® SA8295P | 144.619 ms | 0 - 532 MB | NPU | Depth-Anything-V2 | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 53.383 ms | 0 - 503 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 85.959 ms | 2 - 1938 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 74.312 ms | 2 - 2 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Snapdragon® X Elite | 146.785 ms | 2 - 2 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 121.4 ms | 0 - 2245 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 278.642 ms | 2 - 1855 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 151.589 ms | 2 - 5 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Qualcomm® SA8775P | 150.435 ms | 2 - 1855 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 157.344 ms | 1 - 4 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Qualcomm® SA7255P | 278.642 ms | 2 - 1855 MB | NPU | Depth-Anything-V2 | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 98.782 ms | 2 - 1910 MB | NPU ## License * The license for the original implementation of Depth-Anything-V2 can be found [here](https://github.com/huggingface/transformers/blob/main/LICENSE). ## References * [Depth Anything V2](https://arxiv.org/abs/2406.09414) * [Source Model Implementation](https://github.com/huggingface/transformers/tree/main/src/transformers/models/depth_anything) ## 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).