Hugging Face's logo

qualcomm
/
DDColor

Image-to-Image
PyTorch
LiteRT
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
e6bd825
·
verified ·
1 Parent(s): aabc3e1

v0.38.0

Browse files

See https://github.com/quic/ai-hub-models/releases/v0.38.0 for changelog.

Files changed (10) hide show
  1. .gitattributes +2 -0
  2. DDColor_float.dlc +3 -0
  3. DDColor_float.onnx.zip +3 -0
  4. DDColor_float.tflite +3 -0
  5. DDColor_w8a8.onnx.zip +3 -0
  6. DDColor_w8a8.tflite +3 -0
  7. DEPLOYMENT_MODEL_LICENSE.pdf +3 -0
  8. LICENSE +2 -0
  9. README.md +272 -0
  10. tool-versions.yaml +4 -0
.gitattributes CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ DDColor_float.dlc filter=lfs diff=lfs merge=lfs -text
37
+ DEPLOYMENT_MODEL_LICENSE.pdf filter=lfs diff=lfs merge=lfs -text
DDColor_float.dlc ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02ff60c235b78cca8208a73170f5eb9bd7f9aaab8e02c32b21b0f170bc47460c
3
+ size 225982228
DDColor_float.onnx.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bb5e05a8f00ddb7d5acc47ef165f80d1056ae497ab3a2232423a2e9caba3d9f
3
+ size 206443876
DDColor_float.tflite ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f5877143aed72e501726b13b8a6727a88fd82c9464e705c5f2838ac48b0f191
3
+ size 225521268
DDColor_w8a8.onnx.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:287ed8031f01531675ca5e09241acd95fcfb2c44be9106f551bc69490a4dd7b7
3
+ size 164761245
DDColor_w8a8.tflite ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18285bafea35525b6206d9c2ce93378c7d81e4a3e627a632301d233b9d29f0c8
3
+ size 57450520
DEPLOYMENT_MODEL_LICENSE.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4409f93b0e82531303b3e10f52f1fdfb56467a25f05b7441c6bbd8bb8a64b42c
3
+ size 109629
LICENSE ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ The license of the original trained model can be found at https://github.com/piddnad/DDColor/blob/master/LICENSE.
2
+ The license for the deployable model files (.tflite, .onnx, .dlc, .bin, etc.) can be found in DEPLOYMENT_MODEL_LICENSE.pdf.
README.md ADDED
@@ -0,0 +1,272 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ license: other
4
+ tags:
5
+ - android
6
+ pipeline_tag: image-to-image
7
+
8
+ ---
9
+
10
+ ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ddcolor/web-assets/model_demo.png)
11
+
12
+ # DDColor: Optimized for Mobile Deployment
13
+ ## Colorize image from the black-and-white image
14
+
15
+
16
+ DDColor is a coloring algorithm that produces natural, vivid color results from incoming black and white images.
17
+
18
+ This model is an implementation of DDColor found [here](https://github.com/piddnad/DDColor/).
19
+
20
+
21
+ This repository provides scripts to run DDColor on Qualcomm® devices.
22
+ More details on model performance across various devices, can be found
23
+ [here](https://aihub.qualcomm.com/models/ddcolor).
24
+
25
+
26
+
27
+ ### Model Details
28
+
29
+ - **Model Type:** Model_use_case.image_editing
30
+ - **Model Stats:**
31
+ - Model checkpoint: ddcolor_paper_tiny.pth
32
+ - Input resolution: 224x224
33
+ - Number of parameters: 56.3M
34
+ - Model size (float): 215 MB
35
+ - Model size (w8a8): 54.8 MB
36
+
37
+ | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
38
+ |---|---|---|---|---|---|---|---|---|
39
+ | DDColor | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 277.458 ms | 0 - 365 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
40
+ | DDColor | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 183.714 ms | 1 - 305 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
41
+ | DDColor | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 1248.776 ms | 1 - 257 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
42
+ | DDColor | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 171.409 ms | 0 - 46 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
43
+ | DDColor | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1118.851 ms | 0 - 44 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
44
+ | DDColor | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 1118.703 ms | 0 - 167 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.onnx.zip) |
45
+ | DDColor | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 175.104 ms | 1 - 364 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
46
+ | DDColor | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 1111.959 ms | 0 - 650 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
47
+ | DDColor | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 277.458 ms | 0 - 365 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
48
+ | DDColor | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 173.301 ms | 0 - 43 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
49
+ | DDColor | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 1121.61 ms | 0 - 42 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
50
+ | DDColor | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 190.875 ms | 0 - 262 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
51
+ | DDColor | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 1240.178 ms | 1 - 262 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
52
+ | DDColor | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 170.399 ms | 0 - 46 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
53
+ | DDColor | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 1125.275 ms | 0 - 42 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
54
+ | DDColor | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 175.104 ms | 1 - 364 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
55
+ | DDColor | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 1111.959 ms | 0 - 650 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
56
+ | DDColor | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 123.876 ms | 1 - 384 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
57
+ | DDColor | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 841.212 ms | 1 - 1063 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
58
+ | DDColor | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 846.408 ms | 1 - 370 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.onnx.zip) |
59
+ | DDColor | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 100.104 ms | 1 - 361 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.tflite) |
60
+ | DDColor | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 846.703 ms | 148 - 745 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
61
+ | DDColor | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 865.453 ms | 1 - 337 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.onnx.zip) |
62
+ | DDColor | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1160.679 ms | 142 - 142 MB | NPU | [DDColor.dlc](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.dlc) |
63
+ | DDColor | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1163.214 ms | 113 - 113 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor.onnx.zip) |
64
+ | DDColor | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 3114.062 ms | 0 - 251 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
65
+ | DDColor | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 3537.092 ms | 0 - 283 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
66
+ | DDColor | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1645.578 ms | 0 - 31 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
67
+ | DDColor | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 2894.452 ms | 95 - 314 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.onnx.zip) |
68
+ | DDColor | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1666.302 ms | 0 - 252 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
69
+ | DDColor | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | TFLITE | 585.134 ms | 94 - 398 MB | CPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
70
+ | DDColor | w8a8 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | ONNX | 645.245 ms | 166 - 190 MB | CPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.onnx.zip) |
71
+ | DDColor | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 814.334 ms | 62 - 99 MB | CPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
72
+ | DDColor | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 622.076 ms | 287 - 344 MB | CPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.onnx.zip) |
73
+ | DDColor | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 3114.062 ms | 0 - 251 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
74
+ | DDColor | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 1647.516 ms | 0 - 25 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
75
+ | DDColor | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 4014.446 ms | 0 - 267 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
76
+ | DDColor | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 1645.3 ms | 0 - 23 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
77
+ | DDColor | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1666.302 ms | 0 - 252 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
78
+ | DDColor | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 1248.823 ms | 0 - 263 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
79
+ | DDColor | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 2145.344 ms | 111 - 918 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.onnx.zip) |
80
+ | DDColor | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 1038.25 ms | 0 - 250 MB | NPU | [DDColor.tflite](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.tflite) |
81
+ | DDColor | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 1857.94 ms | 115 - 530 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.onnx.zip) |
82
+ | DDColor | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2988.582 ms | 199 - 199 MB | NPU | [DDColor.onnx.zip](https://huggingface.co/qualcomm/DDColor/blob/main/DDColor_w8a8.onnx.zip) |
83
+
84
+
85
+
86
+
87
+ ## Installation
88
+
89
+
90
+ Install the package via pip:
91
+ ```bash
92
+ pip install "qai-hub-models[ddcolor]"
93
+ ```
94
+
95
+
96
+ ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
97
+
98
+ Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
99
+ Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.
100
+
101
+ With this API token, you can configure your client to run models on the cloud
102
+ hosted devices.
103
+ ```bash
104
+ qai-hub configure --api_token API_TOKEN
105
+ ```
106
+ Navigate to [docs](https://app.aihub.qualcomm.com/docs/) for more information.
107
+
108
+
109
+
110
+ ## Demo off target
111
+
112
+ The package contains a simple end-to-end demo that downloads pre-trained
113
+ weights and runs this model on a sample input.
114
+
115
+ ```bash
116
+ python -m qai_hub_models.models.ddcolor.demo
117
+ ```
118
+
119
+ The above demo runs a reference implementation of pre-processing, model
120
+ inference, and post processing.
121
+
122
+ **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
123
+ environment, please add the following to your cell (instead of the above).
124
+ ```
125
+ %run -m qai_hub_models.models.ddcolor.demo
126
+ ```
127
+
128
+
129
+ ### Run model on a cloud-hosted device
130
+
131
+ In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
132
+ device. This script does the following:
133
+ * Performance check on-device on a cloud-hosted device
134
+ * Downloads compiled assets that can be deployed on-device for Android.
135
+ * Accuracy check between PyTorch and on-device outputs.
136
+
137
+ ```bash
138
+ python -m qai_hub_models.models.ddcolor.export
139
+ ```
140
+
141
+
142
+
143
+ ## How does this work?
144
+
145
+ This [export script](https://aihub.qualcomm.com/models/ddcolor/qai_hub_models/models/DDColor/export.py)
146
+ leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
147
+ on-device. Lets go through each step below in detail:
148
+
149
+ Step 1: **Compile model for on-device deployment**
150
+
151
+ To compile a PyTorch model for on-device deployment, we first trace the model
152
+ in memory using the `jit.trace` and then call the `submit_compile_job` API.
153
+
154
+ ```python
155
+ import torch
156
+
157
+ import qai_hub as hub
158
+ from qai_hub_models.models.ddcolor import Model
159
+
160
+ # Load the model
161
+ torch_model = Model.from_pretrained()
162
+
163
+ # Device
164
+ device = hub.Device("Samsung Galaxy S25")
165
+
166
+ # Trace model
167
+ input_shape = torch_model.get_input_spec()
168
+ sample_inputs = torch_model.sample_inputs()
169
+
170
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
171
+
172
+ # Compile model on a specific device
173
+ compile_job = hub.submit_compile_job(
174
+ model=pt_model,
175
+ device=device,
176
+ input_specs=torch_model.get_input_spec(),
177
+ )
178
+
179
+ # Get target model to run on-device
180
+ target_model = compile_job.get_target_model()
181
+
182
+ ```
183
+
184
+
185
+ Step 2: **Performance profiling on cloud-hosted device**
186
+
187
+ After compiling models from step 1. Models can be profiled model on-device using the
188
+ `target_model`. Note that this scripts runs the model on a device automatically
189
+ provisioned in the cloud. Once the job is submitted, you can navigate to a
190
+ provided job URL to view a variety of on-device performance metrics.
191
+ ```python
192
+ profile_job = hub.submit_profile_job(
193
+ model=target_model,
194
+ device=device,
195
+ )
196
+
197
+ ```
198
+
199
+ Step 3: **Verify on-device accuracy**
200
+
201
+ To verify the accuracy of the model on-device, you can run on-device inference
202
+ on sample input data on the same cloud hosted device.
203
+ ```python
204
+ input_data = torch_model.sample_inputs()
205
+ inference_job = hub.submit_inference_job(
206
+ model=target_model,
207
+ device=device,
208
+ inputs=input_data,
209
+ )
210
+ on_device_output = inference_job.download_output_data()
211
+
212
+ ```
213
+ With the output of the model, you can compute like PSNR, relative errors or
214
+ spot check the output with expected output.
215
+
216
+ **Note**: This on-device profiling and inference requires access to Qualcomm®
217
+ AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
218
+
219
+
220
+
221
+ ## Run demo on a cloud-hosted device
222
+
223
+ You can also run the demo on-device.
224
+
225
+ ```bash
226
+ python -m qai_hub_models.models.ddcolor.demo --eval-mode on-device
227
+ ```
228
+
229
+ **NOTE**: If you want running in a Jupyter Notebook or Google Colab like
230
+ environment, please add the following to your cell (instead of the above).
231
+ ```
232
+ %run -m qai_hub_models.models.ddcolor.demo -- --eval-mode on-device
233
+ ```
234
+
235
+
236
+ ## Deploying compiled model to Android
237
+
238
+
239
+ The models can be deployed using multiple runtimes:
240
+ - TensorFlow Lite (`.tflite` export): [This
241
+ tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
242
+ guide to deploy the .tflite model in an Android application.
243
+
244
+
245
+ - QNN (`.so` export ): This [sample
246
+ app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
247
+ provides instructions on how to use the `.so` shared library in an Android application.
248
+
249
+
250
+ ## View on Qualcomm® AI Hub
251
+ Get more details on DDColor's performance across various devices [here](https://aihub.qualcomm.com/models/ddcolor).
252
+ Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
253
+
254
+
255
+ ## License
256
+ * The license for the original implementation of DDColor can be found
257
+ [here](https://github.com/piddnad/DDColor/blob/master/LICENSE).
258
+ * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
259
+
260
+
261
+
262
+ ## References
263
+ * [DDColor: Towards Photo-Realistic Image Colorization via Dual Decoders](https://arxiv.org/abs/2201.03545)
264
+ * [Source Model Implementation](https://github.com/piddnad/DDColor/)
265
+
266
+
267
+
268
+ ## Community
269
+ * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
270
+ * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
271
+
272
+
tool-versions.yaml ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ tool_versions:
2
+ onnx:
3
+ qairt: 2.37.1.250807093845_124904
4
+ onnx_runtime: 1.22.2