mobile-segmentation - GitLite

KO | EN

gitlite — search

#javascript #hacktoberfest #ai #golang #machine-learning #rust #deep-learning #windows #awesome #cli #awesome-list #vue #mcp #openai

mobile-segmentation

(★ 109)

Real-time semantic image segmentation on mobile devices

파일 탐색기

최종 버전 다운로드 (.zip)

- __init__.py
- dense_prediction_cell.py
- dense_prediction_cell_branch5_top1_cityscapes.json
- feature_extractor.py
- preprocess_utils.py
- shufflenet_v2.py
- utils.py
- __init.py
- build_ade20k_data.py
- build_cityscapes_data.py
- build_coco_data.py
- build_data.py
- convert_cityscapes.sh
- download_and_convert_ade20k.sh
- download_and_convert_ms_coco.sh
- segmentation_dataset.py
- __init__.py
- get_dataset_colormap.py
- input_generator.py
- loss.py
- save_annotation.py
- train_utils.py
.editorconfig
.gitignore
.pylintrc
common.py
evaluate.py
export_tflite.py
input_preprocess.py
LICENSE
model.py
README.md
requirements.txt
train.py
visualize.py

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/sercant/mobile-segmentation

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd mobile-segmentation

방금 내려받은 프로젝트 폴더 안으로 이동합니다.

2. Python

쉬움 추천

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Python 3 설치 시 'Add Python to PATH' 옵션을 꼭 체크하세요 (Windows).

pip install -r requirements.txt

requirements.txt 등에 명시된 파이썬 라이브러리를 설치합니다.

python <실행할 파일명>.py # README에서 정확한 실행 파일명을 확인하세요

파이썬 스크립트(또는 모듈)를 실행합니다.

✅ 에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# An efficient solution for semantic segmentation: ShuffleNet V2 with atrous separable convolutions

We present a computationally efficient approach to semantic segmentation, while achieving a high mean intersection over union (mIOU), 70.33% on Cityscapes challenge. The network proposed is capable of running real-time on mobile devices.

Paper: [10.1007/978-3-030-20205-7_4][4]

If you find the code useful for your research, please consider citing us:

```tex
@InProceedings{turkmen2019efficient,
  author    = {Sercan T{\"u}rkmen and Janne Heikkil{\"a}},
  title     = {An Efficient Solution for Semantic Segmentation: {ShuffleNet} V2 with Atrous Separable Convolutions},
  booktitle = {Image Analysis},
  year      = {2019},
  editor    = {Michael Felsberg and Per-Erik Forss{\'e}n and Ida-Maria Sintorn and Jonas Unger},
  volume    = {11482},
  pages     = {41--53},
  address   = {Cham},
  publisher = {Springer International Publishing},
  doi       = {10.1007/978-3-030-20205-7_4},
  isbn      = {978-3-030-20205-7},
  url       = {http://dx.doi.org/10.1007/978-3-030-20205-7_4},
}
```

## Getting ready

1. Add [`tensorflow/models/slim`][3] to your python path in order to run most of the scripts! To do so follow these steps:
   1. Clone or download the [`tensorflow/models/slim`][3] repository to a separate folder.
   2. Add the path to the repository by running the following code: `export PYTHONPATH=path_to_the_cloned_folder/tensorflow_models/research/slim:${PYTHONPATH}`
2. Prepare dataset. Example scripts and code is available under the `dataset` folder. The dataset should be in `tfrecord` format.

## Model zoo

| Checkpoint name                         | Trained on                                          | Uses DPC | Eval OS | Eval scales | Left-right Flip |    mIOU     | File Size |
| --------------------------------------- | --------------------------------------------------- | :--: | :-----: | :---------: | :-------------: | :---------: | --------: |
| [shufflenetv2_basic_cityscapes_67_7][1] | MS COCO 2017* + Cityscapes coarse + Cityscapes fine | No |  16    |   \[1.0\]   |       No        | 67.7% (val) |     4.9MB |
| [shufflenetv2_dpc_cityscapes_71_3][2]   | MS COCO 2017* + Cityscapes coarse + Cityscapes fine | Yes |  16    |   \[1.0\]   |       No        | 71.3% (val) |     6.3MB |

\* Filtered to include only `person`, `car`, `truck`, `bus`, `train`, `motorcycle`, `bicycle`, `stop sign`, `parking meter` classes and samples that contain over 1000 annotated pixels.

## Training

To learn more about the available flags you can check `common.py` and the specific script that you are trying to run (e.g. `train.py`).

### Example training configuration

```sh
python train.py \
    --model_variant=shufflenet_v2 \
    --tf_initial_checkpoint=./checkpoints/model.ckpt \
    --training_number_of_steps=120000 \
    --base_learning_rate=0.001 \
    --fine_tune_batch_norm=True \
    --initialize_last_layer=False \
    --output_stride=16 \
    --train_crop_size=769 \
    --train_crop_size=769 \
    --train_batch_size=16 \
    --dataset=cityscapes \
    --train_split=train \
    --dataset_dir=./dataset/cityscapes/tfrecord \
    --train_logdir=./logs \
    --loss_function=sce
```

**Important:** To use DPC architecture in your model, you should also set this parameter:

--dense_prediction_cell_json=./core/dense_prediction_cell_branch5_top1_cityscapes.json

### Example evaluation configuration

```sh
python evaluate.py \
    --model_variant=shufflenet_v2 \
    --eval_crop_size=1025 \
    --eval_crop_size=2049 \
    --output_stride=16 \
    --eval_logdir=./logs/eval \
    --checkpoint_dir=./logs \
    --dataset=cityscapes \
    --dataset_dir=./dataset/cityscapes/tfrecord
```

**Important:** If you are trying to evaluate a checkpoint that uses DPC architecture, you should also set this parameter:

--dense_prediction_cell_json=./core/dense_prediction_cell_branch5_top1_cityscapes.json

## Exporting to TFLITE model

`export_tflite.py` script contains several parameters at the top of the script.

## Running on Android

You can find an example script to run the this model and Tensorflow Lite interpreter for segmentation on Android in [this repository][5] and also an example application in [sercant/android-segmentation-app][6].

[1]: https://github.com/sercant/mobile-segmentation/releases/download/v0.1.0/shufflenetv2_basic_cityscapes_67_7.zip
[2]: https://github.com/sercant/mobile-segmentation/releases/download/v0.1.0/shufflenetv2_dpc_cityscapes_71_3.zip
[3]: https://github.com/tensorflow/models/tree/v1.13.0/research/slim
[4]: https://doi.org/10.1007/978-3-030-20205-7_4
[5]: https://github.com/sercant/android-segmentation
[6]: https://github.com/sercant/android-segmentation-app