【人脸属性】Age-Gender-Estimate-TF复现步骤

心已赠人 2022-12-27 08:34 168阅读 0赞

[https://github.com/BoyuanJiang/Age-Gender-Estimate-TF][https_github.com_BoyuanJiang_Age-Gender-Estimate-TF]

# 0.环境 #

ubuntu16.04
    python3.6
    
    tensorflow==1.10.0
    dlib==19.8.1
    opencv-python==4.1.0.25
    matplotlib==3.3.1
    imutils==0.4.3
    numpy==1.19.2
    pandas==1.1.2
    scipy
    scikit-learn
    scikit-image
    tqdm

# 1.准备与处理数据 #

[https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/static/imdb\_crop.tar][https_data.vision.ee.ethz.ch_cvl_rrothe_imdb-wiki_static_imdb_crop.tar]，

[https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/static/wiki\_crop.tar][https_data.vision.ee.ethz.ch_cvl_rrothe_imdb-wiki_static_wiki_crop.tar]。

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70][]

处理数据，首先修改convert\_to\_records\_multiCPU.py，在line 120行替换为如下：

'file_name': _bytes_feature(bytes(file_name[index][0], 'utf-8'))}))

再运行下面的，转换为tfrecord：

python convert_to_records_multiCPU.py --imdb --nworks 8

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 1][]

（这个过程太漫长了，一度以为出错了。7个多小时）

以/data/train目录为例：

![20201209165226278.png][]

![20201209165126643.png][]

# 2.准备预训练模型 #

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 2][]

[https://pan.baidu.com/s/1dFewgqH][https_pan.baidu.com_s_1dFewgqH]（大概80多M），按照作者的要求，放到models目录下。

cp ./weight_from_facenet_20170512-110547.zip ./models/
    cd models/ && unzip ./weight_from_facenet_20170512-110547.zip && cd ..

# 3.训练 #

CUDA_VISIBLE_DEVICES=1 python train.py --lr 1e-3 --weight_decay 1e-5 --epoch 6 --batch_size 128 --keep_prob 0.8 --cuda

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 3][]

这是作者复现的tensorboard可视化结果。

# 4.测试 #

测试部分作者写得很详细：

[https://github.com/BoyuanJiang/Age-Gender-Estimate-TF][https_github.com_BoyuanJiang_Age-Gender-Estimate-TF]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 4][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 5][]

[https_github.com_BoyuanJiang_Age-Gender-Estimate-TF]: https://github.com/BoyuanJiang/Age-Gender-Estimate-TF
[https_data.vision.ee.ethz.ch_cvl_rrothe_imdb-wiki_static_imdb_crop.tar]: https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/static/imdb_crop.tar
[https_data.vision.ee.ethz.ch_cvl_rrothe_imdb-wiki_static_wiki_crop.tar]: https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/static/wiki_crop.tar
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70]: /images/20221120/e09c8ca530444cc6aaca170a29fd2977.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 1]: /images/20221120/19132b7a0f064e31bf91f00b1c55451a.png
[20201209165226278.png]: /images/20221120/eef1aa2b2b40439ebd9fa265b301efb5.png
[20201209165126643.png]: /images/20221120/eb0bcecc76b94e7a9e0446faa6a80a06.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 2]: /images/20221120/f599ed6bf14d43d78ab10d4d5638bbe1.png
[https_pan.baidu.com_s_1dFewgqH]: https://pan.baidu.com/s/1dFewgqH
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 3]: /images/20221120/1cd0ae1ac7654d8bbd981cecba80eb0e.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 4]: /images/20221120/c4abcb189c484389b69e2fac70e2a49a.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM1OTc1NDQ3_size_16_color_FFFFFF_t_70 5]: /images/20221120/68644fcff1584abf98d9cef7bc35a61b.png