Hugging Face's logo

qualcomm
/
DeepLabXception

Image Segmentation
PyTorch
bu_auto
android
Model card Files
xet
 Community
DeepLabXception
File size: 13,280 Bytes
0d3a78b
 
 
 
aba40f5
0d3a78b
 
 
 
 
 
 
69e2e85
0d3a78b
 
 
69e2e85
8d8972f
69e2e85
 
 
 
 
 
 
 
 
 
 
 
8d8972f
 
 
 
 
 
69e2e85
 
 
 
 
 
8d8972f
69e2e85
 
 
 
 
 
8d8972f
69e2e85
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8d8972f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
052606d
8d8972f
 
 
 
 
 
 
 
 
 
052606d
8d8972f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0d3a78b
 
 
 
 
 
 
 
 
 
 
 
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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
---
library_name: pytorch
license: other
tags:
- bu_auto
- android
pipeline_tag: image-segmentation

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deeplab_xception/web-assets/model_demo.png)

# DeepLabXception: Optimized for Qualcomm Devices

DeepLabXception is a semantic segmentation model supporting multiple backbones like ResNet-101 and Xception, with flexible dataset compatibility including COCO, VOC, and Cityscapes.

This is based on the implementation of DeepLabXception found [here](https://github.com/LikeLy-Journey/SegmenTron).
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/deeplab_xception) 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/deeplab_xception/releases/v0.59.0/deeplab_xception-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deeplab_xception/releases/v0.59.0/deeplab_xception-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deeplab_xception/releases/v0.59.0/deeplab_xception-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deeplab_xception/releases/v0.59.0/deeplab_xception-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deeplab_xception/releases/v0.59.0/deeplab_xception-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/deeplab_xception/releases/v0.59.0/deeplab_xception-tflite-w8a8.zip)

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


### 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/deeplab_xception) 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 [DeepLabXception on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/deeplab_xception) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: COCO_WITH_VOC_LABELS_V1
- Input resolution: 480x520
- Number of output classes: 21
- Number of parameters: 41.26M
- Model size (float): 158 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| DeepLabXception | ONNX | float | Snapdragon® X2 Elite | 10.416 ms | 3 - 3 MB | NPU
| DeepLabXception | ONNX | float | Snapdragon® X Elite | 23.036 ms | 86 - 86 MB | NPU
| DeepLabXception | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 17.073 ms | 0 - 356 MB | NPU
| DeepLabXception | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 49.996 ms | 4 - 356 MB | NPU
| DeepLabXception | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 23.03 ms | 0 - 95 MB | NPU
| DeepLabXception | ONNX | float | Qualcomm® QCS8450 | 49.996 ms | 4 - 356 MB | NPU
| DeepLabXception | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 34.389 ms | 3 - 8 MB | NPU
| DeepLabXception | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 23.036 ms | 86 - 86 MB | NPU
| DeepLabXception | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 13.114 ms | 2 - 249 MB | NPU
| DeepLabXception | ONNX | float | Snapdragon® 8 Elite Mobile | 13.114 ms | 2 - 249 MB | NPU
| DeepLabXception | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.905 ms | 2 - 255 MB | NPU
| DeepLabXception | ONNX | w8a8 | Snapdragon® X2 Elite | 3.4 ms | 2 - 2 MB | NPU
| DeepLabXception | ONNX | w8a8 | Snapdragon® X Elite | 7.732 ms | 44 - 44 MB | NPU
| DeepLabXception | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 5.283 ms | 1 - 388 MB | NPU
| DeepLabXception | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 12.702 ms | 0 - 377 MB | NPU
| DeepLabXception | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 45.874 ms | 0 - 4 MB | NPU
| DeepLabXception | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.449 ms | 0 - 50 MB | NPU
| DeepLabXception | ONNX | w8a8 | Qualcomm® QCS8450 | 12.702 ms | 0 - 377 MB | NPU
| DeepLabXception | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 7.332 ms | 0 - 4 MB | NPU
| DeepLabXception | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 7.732 ms | 44 - 44 MB | NPU
| DeepLabXception | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 4.218 ms | 0 - 288 MB | NPU
| DeepLabXception | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 4.218 ms | 0 - 288 MB | NPU
| DeepLabXception | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 3.167 ms | 0 - 298 MB | NPU
| DeepLabXception | QNN_DLC | float | Snapdragon® X2 Elite | 11.523 ms | 3 - 3 MB | NPU
| DeepLabXception | QNN_DLC | float | Snapdragon® X Elite | 22.968 ms | 3 - 3 MB | NPU
| DeepLabXception | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 16.035 ms | 1 - 357 MB | NPU
| DeepLabXception | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 45.83 ms | 3 - 355 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 112.936 ms | 2 - 245 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 22.014 ms | 3 - 5 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® SA8775P | 30.624 ms | 1 - 245 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® SA8650P | 30.624 ms | 1 - 245 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® SA8255P | 30.624 ms | 1 - 245 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® QCS8450 | 45.83 ms | 3 - 355 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 30.941 ms | 3 - 8 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 22.968 ms | 3 - 3 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 12.355 ms | 3 - 252 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® SA7255P | 112.936 ms | 2 - 245 MB | NPU
| DeepLabXception | QNN_DLC | float | Qualcomm® SA8295P | 39.132 ms | 0 - 239 MB | NPU
| DeepLabXception | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 12.355 ms | 3 - 252 MB | NPU
| DeepLabXception | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 9.651 ms | 3 - 260 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 3.887 ms | 1 - 1 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® X Elite | 8.573 ms | 1 - 1 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 5.539 ms | 1 - 372 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 13.853 ms | 1 - 368 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 46.191 ms | 2 - 5 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 20.371 ms | 1 - 283 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.87 ms | 1 - 3 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® SA8775P | 8.058 ms | 1 - 286 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® SA8650P | 8.058 ms | 1 - 286 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® SA8255P | 8.058 ms | 1 - 286 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 13.853 ms | 1 - 368 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 7.763 ms | 1 - 3 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 8.573 ms | 1 - 1 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 93.779 ms | 1 - 303 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 12.454 ms | 1 - 277 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 4.364 ms | 1 - 291 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® SA7255P | 20.371 ms | 1 - 283 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Qualcomm® SA8295P | 12.733 ms | 0 - 279 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 4.364 ms | 1 - 291 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 3.283 ms | 1 - 281 MB | NPU
| DeepLabXception | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 12.454 ms | 1 - 277 MB | NPU
| DeepLabXception | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 16.2 ms | 0 - 455 MB | NPU
| DeepLabXception | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 45.527 ms | 0 - 455 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 113.216 ms | 1 - 343 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 21.435 ms | 0 - 3 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® SA8775P | 30.605 ms | 0 - 343 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® SA8650P | 30.605 ms | 0 - 343 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® SA8255P | 30.605 ms | 0 - 343 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® QCS8450 | 45.527 ms | 0 - 455 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 30.823 ms | 0 - 93 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 12.334 ms | 0 - 353 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® SA7255P | 113.216 ms | 1 - 343 MB | NPU
| DeepLabXception | TFLITE | float | Qualcomm® SA8295P | 38.738 ms | 0 - 336 MB | NPU
| DeepLabXception | TFLITE | float | Snapdragon® 8 Elite Mobile | 12.334 ms | 0 - 353 MB | NPU
| DeepLabXception | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 9.286 ms | 0 - 353 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 5.0 ms | 0 - 368 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 12.048 ms | 0 - 360 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 44.459 ms | 0 - 58 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 19.018 ms | 0 - 284 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 6.933 ms | 0 - 3 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® SA8775P | 7.27 ms | 0 - 284 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® SA8650P | 7.27 ms | 0 - 284 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® SA8255P | 7.27 ms | 0 - 284 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® QCS8450 | 12.048 ms | 0 - 360 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 7.738 ms | 0 - 47 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 100.32 ms | 0 - 291 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 12.071 ms | 0 - 266 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 3.901 ms | 0 - 275 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® SA7255P | 19.018 ms | 0 - 284 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Qualcomm® SA8295P | 11.614 ms | 0 - 276 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 3.901 ms | 0 - 275 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 2.91 ms | 0 - 286 MB | NPU
| DeepLabXception | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 12.071 ms | 0 - 266 MB | NPU

## License
* The license for the original implementation of DeepLabXception can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).

## References
* [Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1802.02611)
* [Source Model Implementation](https://github.com/LikeLy-Journey/SegmenTron)

## 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).