Depth Estimation
PyTorch
android
qaihm-bot commited on
Commit
4c6ba8d
·
verified ·
1 Parent(s): 7519e25

See https://github.com/qualcomm/ai-hub-models/releases/v0.61.0 for changelog.

Files changed (2) hide show
  1. README.md +52 -52
  2. release_assets.json +4 -4
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  library_name: pytorch
3
- license: other
4
  tags:
5
  - android
6
  pipeline_tag: depth-estimation
@@ -14,7 +14,7 @@ pipeline_tag: depth-estimation
14
  Depth Anything 3 (DA3), a model that predicts spatially consistent geometry from arbitrary visual inputs, with or without known camera poses.
15
 
16
  This is based on the implementation of Depth-Anything-V3 found [here](https://github.com/ByteDance-Seed/Depth-Anything-3).
17
- 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/v0.60.0/src/qai_hub_models/models/depth_anything_v3) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
18
 
19
  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.
20
 
@@ -27,23 +27,23 @@ Below are pre-exported model assets ready for deployment.
27
 
28
  | Runtime | Precision | Chipset | SDK Versions | Download |
29
  |---|---|---|---|---|
30
- | 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_v3/releases/v0.60.0/depth_anything_v3-onnx-float.zip)
31
- | QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.60.0/depth_anything_v3-qnn_dlc-float.zip)
32
- | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.60.0/depth_anything_v3-tflite-float.zip)
33
 
34
  For more device-specific assets and performance metrics, visit **[Depth-Anything-V3 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/depth_anything_v3)**.
35
 
36
 
37
  ### Option 2: Export with Custom Configurations
38
 
39
- Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/depth_anything_v3) Python library to compile and export the model with your own:
40
  - Custom weights (e.g., fine-tuned checkpoints)
41
  - Custom input shapes
42
  - Target device and runtime configurations
43
 
44
  This option is ideal if you need to customize the model beyond the default configuration provided here.
45
 
46
- See our repository for [Depth-Anything-V3 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/depth_anything_v3) for usage instructions.
47
 
48
  ## Model Details
49
 
@@ -58,51 +58,51 @@ See our repository for [Depth-Anything-V3 on GitHub](https://github.com/qualcomm
58
  ## Performance Summary
59
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
60
  |---|---|---|---|---|---|---
61
- | Depth-Anything-V3 | ONNX | float | Snapdragon® X2 Elite | 27.824 ms | 5 - 5 MB | NPU
62
- | Depth-Anything-V3 | ONNX | float | Snapdragon® X Elite | 70.34 ms | 70 - 70 MB | NPU
63
- | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 55.074 ms | 2 - 652 MB | NPU
64
- | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 129.049 ms | 0 - 659 MB | NPU
65
- | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 75.942 ms | 3 - 10 MB | NPU
66
- | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 69.067 ms | 4 - 7 MB | NPU
67
- | Depth-Anything-V3 | ONNX | float | Qualcomm® QCS8450 | 129.049 ms | 0 - 659 MB | NPU
68
- | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 77.071 ms | 3 - 9 MB | NPU
69
- | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 70.34 ms | 70 - 70 MB | NPU
70
- | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 36.983 ms | 2 - 500 MB | NPU
71
- | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Elite Mobile | 36.983 ms | 2 - 500 MB | NPU
72
- | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 28.378 ms | 2 - 522 MB | NPU
73
- | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® X2 Elite | 35.145 ms | 3 - 3 MB | NPU
74
- | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® X Elite | 79.04 ms | 3 - 3 MB | NPU
75
- | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 61.72 ms | 3 - 652 MB | NPU
76
- | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 150.395 ms | 4 - 680 MB | NPU
77
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 86.095 ms | 3 - 10 MB | NPU
78
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 186.541 ms | 2 - 492 MB | NPU
79
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 78.894 ms | 3 - 5 MB | NPU
80
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8775P | 85.45 ms | 1 - 491 MB | NPU
81
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8650P | 85.45 ms | 1 - 491 MB | NPU
82
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8255P | 85.45 ms | 1 - 491 MB | NPU
83
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® QCS8450 | 150.395 ms | 4 - 680 MB | NPU
84
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 101.571 ms | 3 - 9 MB | NPU
85
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 79.04 ms | 3 - 3 MB | NPU
86
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 43.602 ms | 0 - 504 MB | NPU
87
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA7255P | 186.541 ms | 2 - 492 MB | NPU
88
- | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8295P | 118.921 ms | 0 - 503 MB | NPU
89
- | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 43.602 ms | 0 - 504 MB | NPU
90
- | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 33.432 ms | 3 - 520 MB | NPU
91
- | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 54.297 ms | 1 - 598 MB | NPU
92
- | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 119.029 ms | 0 - 611 MB | NPU
93
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 79.595 ms | 1 - 82 MB | NPU
94
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 180.694 ms | 2 - 467 MB | NPU
95
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 74.875 ms | 1 - 9 MB | NPU
96
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8775P | 81.144 ms | 1 - 467 MB | NPU
97
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8650P | 81.144 ms | 1 - 467 MB | NPU
98
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8255P | 81.144 ms | 1 - 467 MB | NPU
99
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® QCS8450 | 119.029 ms | 0 - 611 MB | NPU
100
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 81.673 ms | 1 - 81 MB | NPU
101
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 39.237 ms | 0 - 473 MB | NPU
102
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA7255P | 180.694 ms | 2 - 467 MB | NPU
103
- | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8295P | 103.444 ms | 1 - 466 MB | NPU
104
- | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Elite Mobile | 39.237 ms | 0 - 473 MB | NPU
105
- | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 30.771 ms | 1 - 488 MB | NPU
106
 
107
  ## License
108
  * The license for the original implementation of Depth-Anything-V3 can be found
 
1
  ---
2
  library_name: pytorch
3
+ license: apache-2.0
4
  tags:
5
  - android
6
  pipeline_tag: depth-estimation
 
14
  Depth Anything 3 (DA3), a model that predicts spatially consistent geometry from arbitrary visual inputs, with or without known camera poses.
15
 
16
  This is based on the implementation of Depth-Anything-V3 found [here](https://github.com/ByteDance-Seed/Depth-Anything-3).
17
+ 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/v0.61.0/src/qai_hub_models/models/depth_anything_v3) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
18
 
19
  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.
20
 
 
27
 
28
  | Runtime | Precision | Chipset | SDK Versions | Download |
29
  |---|---|---|---|---|
30
+ | 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_v3/releases/v0.61.0/depth_anything_v3-onnx-float.zip)
31
+ | QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.61.0/depth_anything_v3-qnn_dlc-float.zip)
32
+ | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.61.0/depth_anything_v3-tflite-float.zip)
33
 
34
  For more device-specific assets and performance metrics, visit **[Depth-Anything-V3 on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/depth_anything_v3)**.
35
 
36
 
37
  ### Option 2: Export with Custom Configurations
38
 
39
+ Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.61.0/src/qai_hub_models/models/depth_anything_v3) Python library to compile and export the model with your own:
40
  - Custom weights (e.g., fine-tuned checkpoints)
41
  - Custom input shapes
42
  - Target device and runtime configurations
43
 
44
  This option is ideal if you need to customize the model beyond the default configuration provided here.
45
 
46
+ See our repository for [Depth-Anything-V3 on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.61.0/src/qai_hub_models/models/depth_anything_v3) for usage instructions.
47
 
48
  ## Model Details
49
 
 
58
  ## Performance Summary
59
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
60
  |---|---|---|---|---|---|---
61
+ | Depth-Anything-V3 | ONNX | float | Snapdragon® X2 Elite | 27.82 ms | 5 - 5 MB | NPU
62
+ | Depth-Anything-V3 | ONNX | float | Snapdragon® X Elite | 70.382 ms | 71 - 71 MB | NPU
63
+ | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 54.857 ms | 0 - 651 MB | NPU
64
+ | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 129.684 ms | 2 - 658 MB | NPU
65
+ | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 75.617 ms | 3 - 10 MB | NPU
66
+ | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 68.902 ms | 0 - 80 MB | NPU
67
+ | Depth-Anything-V3 | ONNX | float | Qualcomm® QCS8450 | 129.684 ms | 2 - 658 MB | NPU
68
+ | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 76.751 ms | 3 - 9 MB | NPU
69
+ | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 70.382 ms | 71 - 71 MB | NPU
70
+ | Depth-Anything-V3 | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 36.703 ms | 2 - 499 MB | NPU
71
+ | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Elite Mobile | 36.703 ms | 2 - 499 MB | NPU
72
+ | Depth-Anything-V3 | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 28.358 ms | 2 - 522 MB | NPU
73
+ | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® X2 Elite | 35.426 ms | 3 - 3 MB | NPU
74
+ | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® X Elite | 79.346 ms | 3 - 3 MB | NPU
75
+ | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 61.302 ms | 0 - 644 MB | NPU
76
+ | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 150.173 ms | 3 - 679 MB | NPU
77
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 86.087 ms | 3 - 10 MB | NPU
78
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 186.902 ms | 1 - 492 MB | NPU
79
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 79.282 ms | 3 - 5 MB | NPU
80
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8775P | 85.452 ms | 1 - 491 MB | NPU
81
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8650P | 85.452 ms | 1 - 491 MB | NPU
82
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8255P | 85.452 ms | 1 - 491 MB | NPU
83
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® QCS8450 | 150.173 ms | 3 - 679 MB | NPU
84
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 85.874 ms | 3 - 9 MB | NPU
85
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 79.346 ms | 3 - 3 MB | NPU
86
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 42.96 ms | 0 - 501 MB | NPU
87
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA7255P | 186.902 ms | 1 - 492 MB | NPU
88
+ | Depth-Anything-V3 | QNN_DLC | float | Qualcomm® SA8295P | 118.858 ms | 0 - 504 MB | NPU
89
+ | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 42.96 ms | 0 - 501 MB | NPU
90
+ | Depth-Anything-V3 | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 33.443 ms | 2 - 519 MB | NPU
91
+ | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 54.425 ms | 1 - 598 MB | NPU
92
+ | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 118.555 ms | 1 - 612 MB | NPU
93
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 79.426 ms | 1 - 82 MB | NPU
94
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 180.596 ms | 2 - 467 MB | NPU
95
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 75.039 ms | 1 - 12 MB | NPU
96
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8775P | 81.177 ms | 2 - 467 MB | NPU
97
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8650P | 81.177 ms | 2 - 467 MB | NPU
98
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8255P | 81.177 ms | 2 - 467 MB | NPU
99
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® QCS8450 | 118.555 ms | 1 - 612 MB | NPU
100
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 81.844 ms | 1 - 81 MB | NPU
101
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 39.175 ms | 1 - 480 MB | NPU
102
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA7255P | 180.596 ms | 2 - 467 MB | NPU
103
+ | Depth-Anything-V3 | TFLITE | float | Qualcomm® SA8295P | 103.351 ms | 1 - 467 MB | NPU
104
+ | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Elite Mobile | 39.175 ms | 1 - 480 MB | NPU
105
+ | Depth-Anything-V3 | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 30.764 ms | 0 - 488 MB | NPU
106
 
107
  ## License
108
  * The license for the original implementation of Depth-Anything-V3 can be found
release_assets.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "version": "0.60.0",
3
  "precisions": {
4
  "float": {
5
  "universal_assets": {
@@ -8,19 +8,19 @@
8
  "qairt": "2.45.0.260326154327",
9
  "onnx_runtime": "1.27.1"
10
  },
11
- "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.60.0/depth_anything_v3-onnx-float.zip"
12
  },
13
  "qnn_dlc": {
14
  "tool_versions": {
15
  "qairt": "2.45.0.260326154327"
16
  },
17
- "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.60.0/depth_anything_v3-qnn_dlc-float.zip"
18
  },
19
  "tflite": {
20
  "tool_versions": {
21
  "qairt": "2.45.0.260326154327"
22
  },
23
- "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.60.0/depth_anything_v3-tflite-float.zip"
24
  }
25
  }
26
  }
 
1
  {
2
+ "version": "0.61.0",
3
  "precisions": {
4
  "float": {
5
  "universal_assets": {
 
8
  "qairt": "2.45.0.260326154327",
9
  "onnx_runtime": "1.27.1"
10
  },
11
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.61.0/depth_anything_v3-onnx-float.zip"
12
  },
13
  "qnn_dlc": {
14
  "tool_versions": {
15
  "qairt": "2.45.0.260326154327"
16
  },
17
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.61.0/depth_anything_v3-qnn_dlc-float.zip"
18
  },
19
  "tflite": {
20
  "tool_versions": {
21
  "qairt": "2.45.0.260326154327"
22
  },
23
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/depth_anything_v3/releases/v0.61.0/depth_anything_v3-tflite-float.zip"
24
  }
25
  }
26
  }