Hugging Face's logo

qualcomm
/
CREStereo

Depth Estimation
PyTorch
android
Model card Files
xet
 Community
qaihm-bot commited on
Commit
f0f0eda
·
verified ·
1 Parent(s): c6f66d2

v0.60.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.60.0 for changelog.

Files changed (2) hide show
  1. README.md +54 -33
  2. release_assets.json +9 -3
README.md CHANGED
@@ -14,7 +14,7 @@ pipeline_tag: depth-estimation
14
  CREStereo (Cascaded Recurrent Network with Adaptive Correlation) is a CVPR 2022 Oral paper that achieves state-of-the-art stereo matching accuracy.
15
 
16
  This is based on the implementation of CREStereo found [here](https://github.com/ibaiGorordo/CREStereo-Pytorch).
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/v0.59.0/src/qai_hub_models/models/crestereo) 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
 
@@ -27,63 +27,84 @@ 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/crestereo/releases/v0.59.0/crestereo-onnx-float.zip)
31
- | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/crestereo/releases/v0.59.0/crestereo-tflite-float.zip)
 
32
 
33
  For more device-specific assets and performance metrics, visit **[CREStereo on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/crestereo)**.
34
 
35
 
36
  ### Option 2: Export with Custom Configurations
37
 
38
- Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/crestereo) Python library to compile and export the model with your own:
39
  - Custom weights (e.g., fine-tuned checkpoints)
40
  - Custom input shapes
41
  - Target device and runtime configurations
42
 
43
  This option is ideal if you need to customize the model beyond the default configuration provided here.
44
 
45
- See our repository for [CREStereo on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/crestereo) for usage instructions.
46
 
47
  ## Model Details
48
 
49
  **Model Type:** Model_use_case.depth_estimation
50
 
51
  **Model Stats:**
52
- - Model checkpoint: CREStereo ETH3D pretrained (crestereo_eth3d.pt)
53
  - Input: Rectified stereo pair — left and right RGB images
54
  - Input resolution: 240x320
55
- - Output: Disparity map
56
- - Number of parameters: 5.43M
57
  - Model size (float): 20.7 MB
 
 
58
 
59
  ## Performance Summary
60
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
61
  |---|---|---|---|---|---|---
62
- | CREStereo | ONNX | float | Snapdragon® X2 Elite | 91.283 ms | 2 - 2 MB | NPU
63
- | CREStereo | ONNX | float | Snapdragon® X Elite | 211.043 ms | 23 - 23 MB | NPU
64
- | CREStereo | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 147.106 ms | 3 - 1522 MB | NPU
65
- | CREStereo | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 356.457 ms | 4 - 1510 MB | NPU
66
- | CREStereo | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 206.458 ms | 0 - 30 MB | NPU
67
- | CREStereo | ONNX | float | Qualcomm® QCS8450 | 356.457 ms | 4 - 1510 MB | NPU
68
- | CREStereo | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 318.438 ms | 2 - 5 MB | NPU
69
- | CREStereo | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 211.043 ms | 23 - 23 MB | NPU
70
- | CREStereo | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 116.15 ms | 3 - 1161 MB | NPU
71
- | CREStereo | ONNX | float | Snapdragon® 8 Elite Mobile | 116.15 ms | 3 - 1161 MB | NPU
72
- | CREStereo | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 88.4 ms | 1 - 936 MB | NPU
73
- | CREStereo | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 152.812 ms | 1 - 2171 MB | NPU
74
- | CREStereo | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 378.156 ms | 0 - 2318 MB | NPU
75
- | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 813.393 ms | 1 - 1598 MB | NPU
76
- | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 215.767 ms | 0 - 14 MB | NPU
77
- | CREStereo | TFLITE | float | Qualcomm® SA8775P | 281.954 ms | 1 - 1596 MB | NPU
78
- | CREStereo | TFLITE | float | Qualcomm® SA8650P | 281.954 ms | 1 - 1596 MB | NPU
79
- | CREStereo | TFLITE | float | Qualcomm® SA8255P | 281.954 ms | 1 - 1596 MB | NPU
80
- | CREStereo | TFLITE | float | Qualcomm® QCS8450 | 378.156 ms | 0 - 2318 MB | NPU
81
- | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 281.064 ms | 0 - 44 MB | NPU
82
- | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 123.347 ms | 1 - 1653 MB | NPU
83
- | CREStereo | TFLITE | float | Qualcomm® SA7255P | 813.393 ms | 1 - 1598 MB | NPU
84
- | CREStereo | TFLITE | float | Qualcomm® SA8295P | 357.7 ms | 1 - 1858 MB | NPU
85
- | CREStereo | TFLITE | float | Snapdragon® 8 Elite Mobile | 123.347 ms | 1 - 1653 MB | NPU
86
- | CREStereo | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 88.415 ms | 0 - 1271 MB | NPU
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
87
 
88
  ## License
89
  * The license for the original implementation of CREStereo can be found
 
14
  CREStereo (Cascaded Recurrent Network with Adaptive Correlation) is a CVPR 2022 Oral paper that achieves state-of-the-art stereo matching accuracy.
15
 
16
  This is based on the implementation of CREStereo found [here](https://github.com/ibaiGorordo/CREStereo-Pytorch).
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/crestereo) 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/crestereo/releases/v0.60.0/crestereo-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/crestereo/releases/v0.60.0/crestereo-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/crestereo/releases/v0.60.0/crestereo-tflite-float.zip)
33
 
34
  For more device-specific assets and performance metrics, visit **[CREStereo on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/crestereo)**.
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/crestereo) 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 [CREStereo on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/crestereo) for usage instructions.
47
 
48
  ## Model Details
49
 
50
  **Model Type:** Model_use_case.depth_estimation
51
 
52
  **Model Stats:**
 
53
  - Input: Rectified stereo pair — left and right RGB images
54
  - Input resolution: 240x320
55
+ - Model checkpoint: CREStereo ETH3D pretrained (crestereo_eth3d.pt)
 
56
  - Model size (float): 20.7 MB
57
+ - Number of parameters: 5.43M
58
+ - Output: Disparity map
59
 
60
  ## Performance Summary
61
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
62
  |---|---|---|---|---|---|---
63
+ | CREStereo | ONNX | float | Snapdragon® X2 Elite | 87.564 ms | 2 - 2 MB | NPU
64
+ | CREStereo | ONNX | float | Snapdragon® X Elite | 202.736 ms | 21 - 21 MB | NPU
65
+ | CREStereo | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 144.369 ms | 3 - 1334 MB | NPU
66
+ | CREStereo | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 328.168 ms | 3 - 1366 MB | NPU
67
+ | CREStereo | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 271.157 ms | 2 - 6 MB | NPU
68
+ | CREStereo | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 200.076 ms | 0 - 25 MB | NPU
69
+ | CREStereo | ONNX | float | Qualcomm® QCS8450 | 328.168 ms | 3 - 1366 MB | NPU
70
+ | CREStereo | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 269.024 ms | 2 - 5 MB | NPU
71
+ | CREStereo | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 202.736 ms | 21 - 21 MB | NPU
72
+ | CREStereo | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 111.31 ms | 1 - 1112 MB | NPU
73
+ | CREStereo | ONNX | float | Snapdragon® 8 Elite Mobile | 111.31 ms | 1 - 1112 MB | NPU
74
+ | CREStereo | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 79.199 ms | 1 - 837 MB | NPU
75
+ | CREStereo | QNN_DLC | float | Snapdragon® X2 Elite | 88.18 ms | 2 - 2 MB | NPU
76
+ | CREStereo | QNN_DLC | float | Snapdragon® X Elite | 203.778 ms | 2 - 2 MB | NPU
77
+ | CREStereo | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 144.84 ms | 2 - 1242 MB | NPU
78
+ | CREStereo | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 328.828 ms | 2 - 1292 MB | NPU
79
+ | CREStereo | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 271.144 ms | 2 - 7 MB | NPU
80
+ | CREStereo | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 797.535 ms | 3 - 1015 MB | NPU
81
+ | CREStereo | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 199.896 ms | 3 - 6 MB | NPU
82
+ | CREStereo | QNN_DLC | float | Qualcomm® SA8775P | 269.996 ms | 3 - 1059 MB | NPU
83
+ | CREStereo | QNN_DLC | float | Qualcomm® SA8650P | 269.996 ms | 3 - 1059 MB | NPU
84
+ | CREStereo | QNN_DLC | float | Qualcomm® SA8255P | 269.996 ms | 3 - 1059 MB | NPU
85
+ | CREStereo | QNN_DLC | float | Qualcomm® QCS8450 | 328.828 ms | 2 - 1292 MB | NPU
86
+ | CREStereo | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 269.305 ms | 4 - 8 MB | NPU
87
+ | CREStereo | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 203.778 ms | 2 - 2 MB | NPU
88
+ | CREStereo | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 111.55 ms | 3 - 1046 MB | NPU
89
+ | CREStereo | QNN_DLC | float | Qualcomm® SA7255P | 797.535 ms | 3 - 1015 MB | NPU
90
+ | CREStereo | QNN_DLC | float | Qualcomm® SA8295P | 312.046 ms | 3 - 1066 MB | NPU
91
+ | CREStereo | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 111.55 ms | 3 - 1046 MB | NPU
92
+ | CREStereo | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 79.208 ms | 1 - 748 MB | NPU
93
+ | CREStereo | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 151.227 ms | 1 - 1672 MB | NPU
94
+ | CREStereo | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 379.427 ms | 1 - 1643 MB | NPU
95
+ | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 278.444 ms | 0 - 44 MB | NPU
96
+ | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 809.84 ms | 1 - 1402 MB | NPU
97
+ | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 216.12 ms | 0 - 7 MB | NPU
98
+ | CREStereo | TFLITE | float | Qualcomm® SA8775P | 282.462 ms | 1 - 1411 MB | NPU
99
+ | CREStereo | TFLITE | float | Qualcomm® SA8650P | 282.462 ms | 1 - 1411 MB | NPU
100
+ | CREStereo | TFLITE | float | Qualcomm® SA8255P | 282.462 ms | 1 - 1411 MB | NPU
101
+ | CREStereo | TFLITE | float | Qualcomm® QCS8450 | 379.427 ms | 1 - 1643 MB | NPU
102
+ | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 281.541 ms | 0 - 43 MB | NPU
103
+ | CREStereo | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 122.089 ms | 1 - 1377 MB | NPU
104
+ | CREStereo | TFLITE | float | Qualcomm® SA7255P | 809.84 ms | 1 - 1402 MB | NPU
105
+ | CREStereo | TFLITE | float | Qualcomm® SA8295P | 350.245 ms | 1 - 1383 MB | NPU
106
+ | CREStereo | TFLITE | float | Snapdragon® 8 Elite Mobile | 122.089 ms | 1 - 1377 MB | NPU
107
+ | CREStereo | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 88.324 ms | 0 - 1114 MB | NPU
108
 
109
  ## License
110
  * The license for the original implementation of CREStereo can be found
release_assets.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "version": "0.59.0",
3
  "precisions": {
4
  "float": {
5
  "universal_assets": {
@@ -8,13 +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/crestereo/releases/v0.59.0/crestereo-onnx-float.zip"
 
 
 
 
 
 
12
  },
13
  "tflite": {
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/crestereo/releases/v0.59.0/crestereo-tflite-float.zip"
18
  }
19
  }
20
  }
 
1
  {
2
+ "version": "0.60.0",
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/crestereo/releases/v0.60.0/crestereo-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/crestereo/releases/v0.60.0/crestereo-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/crestereo/releases/v0.60.0/crestereo-tflite-float.zip"
24
  }
25
  }
26
  }