Hugging Face's logo

qualcomm
/
DnCNN

Image-to-Image
PyTorch
real_time
android
Model card Files
xet
 Community
qaihm-bot commited on
Commit
29af016
·
verified ·
1 Parent(s): d59af1f

v0.51.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.51.0 for changelog.

Files changed (3) hide show
  1. LICENSE +1 -0
  2. README.md +144 -0
  3. release_assets.json +53 -0
LICENSE ADDED
@@ -0,0 +1 @@
 
 
1
+ The license of the original trained model can be found at https://github.com/cszn/KAIR/blob/master/LICENSE.
README.md ADDED
@@ -0,0 +1,144 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ license: other
4
+ tags:
5
+ - real_time
6
+ - android
7
+ pipeline_tag: image-to-image
8
+
9
+ ---
10
+
11
+ ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/web-assets/model_demo.png)
12
+
13
+ # DnCNN: Optimized for Qualcomm Devices
14
+
15
+ DnCNN is a 17-layer denoising convolutional neural network that uses residual learning to remove Gaussian noise (sigma=25) from grayscale images. The network predicts the noise residual and subtracts it from the input to produce a clean image.
16
+
17
+ This is based on the implementation of DnCNN found [here](https://github.com/cszn/KAIR).
18
+ 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/dncnn) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
+
20
+ 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.
21
+
22
+ ## Getting Started
23
+ There are two ways to deploy this model on your device:
24
+
25
+ ### Option 1: Download Pre-Exported Models
26
+
27
+ Below are pre-exported model assets ready for deployment.
28
+
29
+ | Runtime | Precision | Chipset | SDK Versions | Download |
30
+ |---|---|---|---|---|
31
+ | 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/dncnn/releases/v0.51.0/dncnn-onnx-float.zip)
32
+ | ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-onnx-w8a8.zip)
33
+ | QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-qnn_dlc-float.zip)
34
+ | QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-qnn_dlc-w8a8.zip)
35
+ | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-tflite-float.zip)
36
+ | TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-tflite-w8a8.zip)
37
+
38
+ For more device-specific assets and performance metrics, visit **[DnCNN on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/dncnn)**.
39
+
40
+
41
+ ### Option 2: Export with Custom Configurations
42
+
43
+ Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/dncnn) Python library to compile and export the model with your own:
44
+ - Custom weights (e.g., fine-tuned checkpoints)
45
+ - Custom input shapes
46
+ - Target device and runtime configurations
47
+
48
+ This option is ideal if you need to customize the model beyond the default configuration provided here.
49
+
50
+ See our repository for [DnCNN on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/dncnn) for usage instructions.
51
+
52
+ ## Model Details
53
+
54
+ **Model Type:** Model_use_case.image_editing
55
+
56
+ **Model Stats:**
57
+ - Model checkpoint: dncnn_25
58
+ - Input resolution: 256x256
59
+ - Number of parameters: 555K
60
+ - Model size (float): 2.12 MB
61
+ - Model size (w8a8): 581 KB
62
+
63
+ ## Performance Summary
64
+ | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
65
+ |---|---|---|---|---|---|---
66
+ | DnCNN | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.123 ms | 0 - 142 MB | NPU
67
+ | DnCNN | ONNX | float | Snapdragon® X2 Elite | 4.052 ms | 0 - 0 MB | NPU
68
+ | DnCNN | ONNX | float | Snapdragon® X Elite | 7.387 ms | 0 - 0 MB | NPU
69
+ | DnCNN | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.179 ms | 1 - 181 MB | NPU
70
+ | DnCNN | ONNX | float | Qualcomm® QCS8550 (Proxy) | 6.942 ms | 1 - 3 MB | NPU
71
+ | DnCNN | ONNX | float | Qualcomm® QCS9075 | 14.352 ms | 1 - 4 MB | NPU
72
+ | DnCNN | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.131 ms | 0 - 137 MB | NPU
73
+ | DnCNN | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.752 ms | 0 - 33 MB | NPU
74
+ | DnCNN | ONNX | w8a8 | Snapdragon® X2 Elite | 1.043 ms | 0 - 0 MB | NPU
75
+ | DnCNN | ONNX | w8a8 | Snapdragon® X Elite | 2.02 ms | 0 - 0 MB | NPU
76
+ | DnCNN | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.336 ms | 0 - 48 MB | NPU
77
+ | DnCNN | ONNX | w8a8 | Qualcomm® QCS6490 | 132.023 ms | 60 - 63 MB | CPU
78
+ | DnCNN | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.792 ms | 0 - 27 MB | NPU
79
+ | DnCNN | ONNX | w8a8 | Qualcomm® QCS9075 | 1.933 ms | 0 - 3 MB | NPU
80
+ | DnCNN | ONNX | w8a8 | Qualcomm® QCM6690 | 191.916 ms | 59 - 67 MB | CPU
81
+ | DnCNN | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.213 ms | 0 - 29 MB | NPU
82
+ | DnCNN | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 158.816 ms | 63 - 71 MB | CPU
83
+ | DnCNN | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.963 ms | 0 - 147 MB | NPU
84
+ | DnCNN | QNN_DLC | float | Snapdragon® X2 Elite | 4.193 ms | 0 - 0 MB | NPU
85
+ | DnCNN | QNN_DLC | float | Snapdragon® X Elite | 7.187 ms | 0 - 0 MB | NPU
86
+ | DnCNN | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 5.064 ms | 0 - 174 MB | NPU
87
+ | DnCNN | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 55.989 ms | 0 - 141 MB | NPU
88
+ | DnCNN | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 6.705 ms | 0 - 2 MB | NPU
89
+ | DnCNN | QNN_DLC | float | Qualcomm® SA8775P | 13.88 ms | 0 - 143 MB | NPU
90
+ | DnCNN | QNN_DLC | float | Qualcomm® QCS9075 | 14.106 ms | 2 - 4 MB | NPU
91
+ | DnCNN | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 13.485 ms | 0 - 177 MB | NPU
92
+ | DnCNN | QNN_DLC | float | Qualcomm® SA7255P | 55.989 ms | 0 - 141 MB | NPU
93
+ | DnCNN | QNN_DLC | float | Qualcomm® SA8295P | 15.308 ms | 0 - 139 MB | NPU
94
+ | DnCNN | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.033 ms | 0 - 146 MB | NPU
95
+ | DnCNN | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.754 ms | 0 - 30 MB | NPU
96
+ | DnCNN | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 1.183 ms | 0 - 0 MB | NPU
97
+ | DnCNN | QNN_DLC | w8a8 | Snapdragon® X Elite | 2.015 ms | 0 - 0 MB | NPU
98
+ | DnCNN | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.332 ms | 0 - 46 MB | NPU
99
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 7.573 ms | 0 - 2 MB | NPU
100
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 7.565 ms | 0 - 28 MB | NPU
101
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.788 ms | 0 - 15 MB | NPU
102
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® SA8775P | 2.0 ms | 0 - 30 MB | NPU
103
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 1.929 ms | 0 - 2 MB | NPU
104
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 38.903 ms | 0 - 144 MB | NPU
105
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 2.36 ms | 0 - 47 MB | NPU
106
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® SA7255P | 7.565 ms | 0 - 28 MB | NPU
107
+ | DnCNN | QNN_DLC | w8a8 | Qualcomm® SA8295P | 4.231 ms | 0 - 26 MB | NPU
108
+ | DnCNN | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.212 ms | 0 - 29 MB | NPU
109
+ | DnCNN | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 3.255 ms | 0 - 140 MB | NPU
110
+ | DnCNN | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.089 ms | 0 - 147 MB | NPU
111
+ | DnCNN | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.179 ms | 0 - 178 MB | NPU
112
+ | DnCNN | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 56.396 ms | 0 - 141 MB | NPU
113
+ | DnCNN | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 6.931 ms | 1 - 2 MB | NPU
114
+ | DnCNN | TFLITE | float | Qualcomm® SA8775P | 14.17 ms | 0 - 145 MB | NPU
115
+ | DnCNN | TFLITE | float | Qualcomm® QCS9075 | 14.322 ms | 0 - 4 MB | NPU
116
+ | DnCNN | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 13.926 ms | 1 - 176 MB | NPU
117
+ | DnCNN | TFLITE | float | Qualcomm® SA7255P | 56.396 ms | 0 - 141 MB | NPU
118
+ | DnCNN | TFLITE | float | Qualcomm® SA8295P | 15.622 ms | 0 - 141 MB | NPU
119
+ | DnCNN | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.126 ms | 0 - 143 MB | NPU
120
+ | DnCNN | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.72 ms | 0 - 32 MB | NPU
121
+ | DnCNN | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.297 ms | 0 - 47 MB | NPU
122
+ | DnCNN | TFLITE | w8a8 | Qualcomm® QCS6490 | 7.794 ms | 0 - 3 MB | NPU
123
+ | DnCNN | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 7.461 ms | 0 - 29 MB | NPU
124
+ | DnCNN | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 1.744 ms | 0 - 1 MB | NPU
125
+ | DnCNN | TFLITE | w8a8 | Qualcomm® SA8775P | 1.977 ms | 0 - 31 MB | NPU
126
+ | DnCNN | TFLITE | w8a8 | Qualcomm® QCS9075 | 1.843 ms | 0 - 3 MB | NPU
127
+ | DnCNN | TFLITE | w8a8 | Qualcomm® QCM6690 | 38.948 ms | 0 - 141 MB | NPU
128
+ | DnCNN | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 2.354 ms | 0 - 48 MB | NPU
129
+ | DnCNN | TFLITE | w8a8 | Qualcomm® SA7255P | 7.461 ms | 0 - 29 MB | NPU
130
+ | DnCNN | TFLITE | w8a8 | Qualcomm® SA8295P | 4.179 ms | 0 - 27 MB | NPU
131
+ | DnCNN | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.164 ms | 0 - 29 MB | NPU
132
+ | DnCNN | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 3.214 ms | 0 - 141 MB | NPU
133
+
134
+ ## License
135
+ * The license for the original implementation of DnCNN can be found
136
+ [here](https://github.com/cszn/KAIR/blob/master/LICENSE).
137
+
138
+ ## References
139
+ * [Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising](https://arxiv.org/abs/1608.03981)
140
+ * [Source Model Implementation](https://github.com/cszn/KAIR)
141
+
142
+ ## Community
143
+ * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
144
+ * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
release_assets.json ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": "0.51.0",
3
+ "precisions": {
4
+ "float": {
5
+ "universal_assets": {
6
+ "tflite": {
7
+ "tool_versions": {
8
+ "qairt": "2.45.0.260326154327",
9
+ "litert": "1.4.2"
10
+ },
11
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-tflite-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/dncnn/releases/v0.51.0/dncnn-qnn_dlc-float.zip"
18
+ },
19
+ "onnx": {
20
+ "tool_versions": {
21
+ "qairt": "2.42.0.251225135753_193295",
22
+ "onnx_runtime": "1.24.3"
23
+ },
24
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-onnx-float.zip"
25
+ }
26
+ }
27
+ },
28
+ "w8a8": {
29
+ "universal_assets": {
30
+ "tflite": {
31
+ "tool_versions": {
32
+ "qairt": "2.45.0.260326154327",
33
+ "litert": "1.4.2"
34
+ },
35
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-tflite-w8a8.zip"
36
+ },
37
+ "qnn_dlc": {
38
+ "tool_versions": {
39
+ "qairt": "2.45.0.260326154327"
40
+ },
41
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-qnn_dlc-w8a8.zip"
42
+ },
43
+ "onnx": {
44
+ "tool_versions": {
45
+ "qairt": "2.42.0.251225135753_193295",
46
+ "onnx_runtime": "1.24.3"
47
+ },
48
+ "download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/dncnn/releases/v0.51.0/dncnn-onnx-w8a8.zip"
49
+ }
50
+ }
51
+ }
52
+ }
53
+ }