Hugging Face's logo

qualcomm
/
Depth-Anything

Depth Estimation
PyTorch
android
Model card Files
xet
 Community
Depth-Anything
File size: 10,633 Bytes
9245eb2
 
aceb802
9245eb2
 
4d45364
9245eb2
 
 
 
 
d8fc06a
9245eb2
 
 
d8fc06a
5b1626e
d8fc06a
 
 
 
 
 
 
 
 
 
 
 
5b1626e
 
 
 
 
d8fc06a
 
 
 
 
 
5b1626e
d8fc06a
 
 
 
 
 
5b1626e
d8fc06a
 
 
 
 
 
 
 
 
 
 
 
 
 
5b1626e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9245eb2
 
a066663
 
9245eb2
 
 
 
 
 
 
 
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
---
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/web-assets/model_demo.png)

# Depth-Anything: 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 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/v0.59.0/src/qai_hub_models/models/depth_anything) 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 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything/releases/v0.59.0/depth_anything-onnx-float.zip)
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything/releases/v0.59.0/depth_anything-onnx-w8a16.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/releases/v0.59.0/depth_anything-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/releases/v0.59.0/depth_anything-qnn_dlc-w8a16.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything/releases/v0.59.0/depth_anything-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[Depth-Anything on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/depth_anything)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/depth_anything) 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 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/depth_anything) for usage instructions.

## Model Details

**Model Type:** Model_use_case.depth_estimation

**Model Stats:**
- Model checkpoint: DepthAnything_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 | ONNX | float | Snapdragon® X2 Elite | 19.086 ms | 5 - 5 MB | NPU
| Depth-Anything | ONNX | float | Snapdragon® X Elite | 49.118 ms | 48 - 48 MB | NPU
| Depth-Anything | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 35.642 ms | 0 - 469 MB | NPU
| Depth-Anything | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 86.214 ms | 3 - 480 MB | NPU
| Depth-Anything | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 47.425 ms | 0 - 55 MB | NPU
| Depth-Anything | ONNX | float | Qualcomm® QCS8450 | 86.214 ms | 3 - 480 MB | NPU
| Depth-Anything | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 55.351 ms | 3 - 9 MB | NPU
| Depth-Anything | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 49.118 ms | 48 - 48 MB | NPU
| Depth-Anything | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 24.427 ms | 2 - 366 MB | NPU
| Depth-Anything | ONNX | float | Snapdragon® 8 Elite Mobile | 24.427 ms | 2 - 366 MB | NPU
| Depth-Anything | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 18.494 ms | 2 - 370 MB | NPU
| Depth-Anything | ONNX | w8a16 | Snapdragon® X2 Elite | 12.149 ms | 4 - 4 MB | NPU
| Depth-Anything | ONNX | w8a16 | Snapdragon® X Elite | 27.706 ms | 27 - 27 MB | NPU
| Depth-Anything | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 18.986 ms | 3 - 513 MB | NPU
| Depth-Anything | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 39.084 ms | 3 - 513 MB | NPU
| Depth-Anything | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 107.209 ms | 2 - 5 MB | NPU
| Depth-Anything | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 27.105 ms | 0 - 34 MB | NPU
| Depth-Anything | ONNX | w8a16 | Qualcomm® QCS8450 | 39.084 ms | 3 - 513 MB | NPU
| Depth-Anything | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 27.906 ms | 2 - 5 MB | NPU
| Depth-Anything | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 27.706 ms | 27 - 27 MB | NPU
| Depth-Anything | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 14.139 ms | 3 - 420 MB | NPU
| Depth-Anything | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 14.139 ms | 3 - 420 MB | NPU
| Depth-Anything | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 11.582 ms | 1 - 426 MB | NPU
| Depth-Anything | QNN_DLC | float | Snapdragon® X2 Elite | 20.128 ms | 3 - 3 MB | NPU
| Depth-Anything | QNN_DLC | float | Snapdragon® X Elite | 51.263 ms | 3 - 3 MB | NPU
| Depth-Anything | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 38.593 ms | 2 - 510 MB | NPU
| Depth-Anything | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 92.024 ms | 3 - 485 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 142.253 ms | 1 - 388 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 50.965 ms | 3 - 6 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® SA8775P | 56.655 ms | 1 - 372 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® SA8650P | 56.655 ms | 1 - 372 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® SA8255P | 56.655 ms | 1 - 372 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® QCS8450 | 92.024 ms | 3 - 485 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 56.952 ms | 3 - 9 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 51.263 ms | 3 - 3 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 24.77 ms | 3 - 386 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® SA7255P | 142.253 ms | 1 - 388 MB | NPU
| Depth-Anything | QNN_DLC | float | Qualcomm® SA8295P | 82.581 ms | 0 - 363 MB | NPU
| Depth-Anything | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 24.77 ms | 3 - 386 MB | NPU
| Depth-Anything | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 19.692 ms | 3 - 384 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 70.118 ms | 2 - 2 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Snapdragon® X Elite | 137.8 ms | 2 - 2 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 113.136 ms | 0 - 2299 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8275 | 258.781 ms | 2 - 2002 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 160.356 ms | 2 - 4 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® SA8775P | 141.295 ms | 2 - 2000 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® SA8650P | 141.295 ms | 2 - 2000 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® SA8255P | 141.295 ms | 2 - 2000 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 136.754 ms | 1 - 5 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 137.8 ms | 2 - 2 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 93.522 ms | 2 - 2015 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Qualcomm® SA7255P | 258.781 ms | 2 - 2002 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 93.522 ms | 2 - 2015 MB | NPU
| Depth-Anything | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 81.583 ms | 2 - 2016 MB | NPU
| Depth-Anything | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 38.779 ms | 1 - 518 MB | NPU
| Depth-Anything | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 92.599 ms | 2 - 491 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 142.42 ms | 1 - 396 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 51.5 ms | 1 - 6 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® SA8775P | 57.021 ms | 1 - 395 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® SA8650P | 57.021 ms | 1 - 395 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® SA8255P | 57.021 ms | 1 - 395 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® QCS8450 | 92.599 ms | 2 - 491 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 64.611 ms | 1 - 58 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 24.685 ms | 1 - 395 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® SA7255P | 142.42 ms | 1 - 396 MB | NPU
| Depth-Anything | TFLITE | float | Qualcomm® SA8295P | 82.521 ms | 1 - 370 MB | NPU
| Depth-Anything | TFLITE | float | Snapdragon® 8 Elite Mobile | 24.685 ms | 1 - 395 MB | NPU
| Depth-Anything | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 19.687 ms | 0 - 390 MB | NPU

## License
* The license for the original implementation of Depth-Anything can be found
  [here](https://github.com/huggingface/transformers/blob/main/LICENSE).

## References
* [Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data](https://arxiv.org/abs/2401.10891)
* [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).