model_zoo.md

# TensorFlow DeepLab Model Zoo

We provide deeplab models pretrained on PASCAL VOC 2012 and Cityscapes datasets
for reproducing our results, as well as some checkpoints that are only
pretrained on ImageNet for training your own models.

## DeepLab models trained on PASCAL VOC 2012

Un-tar'ed directory includes:

*   a frozen inference graph (`frozen_inference_graph.pb`). All frozen inference
    graphs use output stride of 8 and a single eval scale of 1.0. No left-right
    flips are used.

*   a checkpoint (`model.ckpt.data-00000-of-00001`, `model.ckpt.index`)

### Model details

We provide several checkpoints that have been pretrained on VOC 2012 train_aug
set or train_aug + trainval set. In the former case, one could train their model
with smaller batch size and freeze batch normalization when limited GPU memory
is available, since we have already fine-tuned the batch normalization for you.
In the latter case, one could directly evaluate the checkpoints on VOC 2012 test
set or use this checkpoint for demo.

Checkpoint name             | Network backbone | Pretrained  dataset | ASPP  | Decoder
--------------------------- | :--------------: | :-----------------: | :---: | :-----:
xception_coco_voc_trainaug  | Xception_65  | MS-COCO <br> VOC 2012 train_aug set| [6,12,18] for OS=16 <br> [12,24,36] for OS=8 | OS = 4
xception_coco_voc_trainval  | Xception_65  | MS-COCO <br> VOC 2012 train_aug + trainval sets | [6,12,18] for OS=16 <br> [12,24,36] for OS=8 | OS = 4

In the table, **OS** denotes output stride.

Checkpoint name                                                                                                          | Eval OS   | Eval scales                | Left-right Flip | Multiply-Adds        | Runtime (sec)  | PASCAL mIOU                    | File Size
------------------------------------------------------------------------------------------------------------------------ | :-------: | :------------------------: | :-------------: | :------------------: | :------------: | :----------------------------: | :-------:
[xception_coco_voc_trainaug](http://download.tensorflow.org/models/deeplabv3_pascal_train_aug_2018_01_04.tar.gz)         | 16 <br> 8 | [1.0] <br> [0.5:0.25:1.75] | No <br> Yes     | 54.17B <br> 3055.35B | 0.7 <br> 223.2 | 82.20% (val) <br> 83.58% (val) | 439MB
[xception_coco_voc_trainval](http://download.tensorflow.org/models/deeplabv3_pascal_trainval_2018_01_04.tar.gz)          | 8         | [0.5:0.25:1.75]            | Yes             | 3055.35B             | 223.2          | 87.80% (**test**)              | 439MB

In the table, we report both computation complexity (in terms of Multiply-Adds
and CPU Runtime) and segmentation performance (in terms of mIOU) on the PASCAL
VOC val or test set. The reported runtime is calculated by tfprof on a
workstation with CPU E5-1650 v3 @ 3.50GHz and 32GB memory. Note that applying
multi-scale inputs and left-right flips increases the segmentation performance
but also significantly increases the computation and thus may not be suitable
for real-time applications.

## DeepLab models trained on Cityscapes

### Model details

We provide several checkpoints that have been pretrained on Cityscapes
train_fine set.

Checkpoint name                       | Network backbone | Pretrained dataset                      | ASPP                                             | Decoder
------------------------------------- | :--------------: | :-------------------------------------: | :----------------------------------------------: | :-----:
xception_cityscapes_trainfine         | Xception_65      | ImageNet <br> Cityscapes train_fine set | [6, 12, 18] for OS=16 <br> [12, 24, 36] for OS=8 | OS = 4

In the table, **OS** denotes output stride.

Checkpoint name                                                                                                                  | Eval OS   | Eval scales                 | Left-right Flip | Multiply-Adds         | Runtime (sec)  | Cityscapes mIOU                | File Size
-------------------------------------------------------------------------------------------------------------------------------- | :-------: | :-------------------------: | :-------------: | :-------------------: | :------------: | :----------------------------: | :-------:
[xception_cityscapes_trainfine](http://download.tensorflow.org/models/deeplabv3_cityscapes_train_2018_02_06.tar.gz)              | 16 <br> 8 | [1.0] <br> [0.75:0.25:1.25] | No <br> Yes     | 418.64B <br> 8677.92B | 5.0 <br> 422.8 | 78.79% (val) <br> 80.42% (val) | 439MB

## Checkpoints pretrained on ImageNet

Un-tar'ed directory includes:

*   model checkpoint (`model.ckpt.data-00000-of-00001`, `model.ckpt.index`).

### Model details

We also provide some checkpoints that are only pretrained on ImageNet so that
one could use this for training your own models.

*   xception: We adapt the original Xception model to the task of semantic
    segmentation with the following changes: (1) more layers, (2) all max
    pooling operations are replaced by strided (atrous) separable convolutions,
    and (3) extra batch-norm and ReLU after each 3x3 depthwise convolution are
    added.

Model name                                                                             | File Size
-------------------------------------------------------------------------------------- | :-------:
[xception](http://download.tensorflow.org/models/deeplabv3_xception_2018_01_04.tar.gz) | 447MB

## References

1.  **Mobilenets: Efficient convolutional neural networks for mobile vision applications**<br />
    Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, Marco Andreetto, Hartwig Adam<br />
    [[link]](https://arxiv.org/abs/1704.04861). arXiv:1704.04861, 2017.

2.  **Inverted Residuals and Linear Bottlenecks: Mobile Networks for Classification, Detection and Segmentation**<br />
    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, Liang-Chieh Chen<br />
    [[link]](https://arxiv.org/abs/1801.04381). arXiv:1801.04381, 2018.

3.  **Xception: Deep Learning with Depthwise Separable Convolutions**<br />
    François Chollet<br />
    [[link]](https://arxiv.org/abs/1610.02357). In the Proc. of CVPR, 2017.

4.  **Deformable Convolutional Networks -- COCO Detection and Segmentation Challenge 2017 Entry**<br />
    Haozhi Qi, Zheng Zhang, Bin Xiao, Han Hu, Bowen Cheng, Yichen Wei, Jifeng Dai<br />
    [[link]](http://presentations.cocodataset.org/COCO17-Detect-MSRA.pdf). ICCV COCO Challenge
    Workshop, 2017.

5.  **The Pascal Visual Object Classes Challenge: A Retrospective**<br />
    Mark Everingham, S. M. Ali Eslami, Luc Van Gool, Christopher K. I. Williams, John M. Winn, Andrew Zisserman<br />
    [[link]](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/). IJCV, 2014.

6.  **Semantic Contours from Inverse Detectors**<br />
    Bharath Hariharan, Pablo Arbelaez, Lubomir Bourdev, Subhransu Maji, Jitendra Malik<br />
    [[link]](http://home.bharathh.info/pubs/codes/SBD/download.html). In the Proc. of ICCV, 2011.

7.  **The Cityscapes Dataset for Semantic Urban Scene Understanding**<br />
    Cordts, Marius, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, Bernt Schiele. <br />
    [[link]](https://www.cityscapes-dataset.com/). In the Proc. of CVPR, 2016.

8.  **Microsoft COCO: Common Objects in Context**<br />
    Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro Perona, Deva Ramanan, C. Lawrence Zitnick, Piotr Dollar<br />
    [[link]](http://cocodataset.org/). In the Proc. of ECCV, 2014.

9.  **ImageNet Large Scale Visual Recognition Challenge**<br />
    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, Li Fei-Fei<br />
    [[link]](http://www.image-net.org/). IJCV, 2015.