基于tf2-keras-yolo3的自定义数据集操作

安装tf2-keras-yolov3环境参考的是 笔墨留年 这位大佬的教程，十分感谢！

自己创建的名为yol3的虚拟环境的库，
强推anaconda+Pycharm
用conda创建虚拟环境，换源，pip下载tensorflow-gpu==2.2.0,
conda安装cudnn和cudatoolkit，一套弄好会减少很多麻烦，不用去自己配置CUDA和cudnn。

absl-py                   0.11.0                    <pip>
astunparse                1.6.3                     <pip>
ca-certificates           2020.12.5            h5b45459_0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
cachetools                4.2.1                     <pip>
certifi                   2020.12.5        py37h03978a9_1    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
chardet                   4.0.0                     <pip>
cudatoolkit               10.1.243             h3826478_8    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
cudnn                     7.6.5.32             h36d860d_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
cycler                    0.10.0                     py_2    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
freeglut                  3.2.1                h0e60522_2    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
freetype                  2.10.4               h546665d_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
gast                      0.3.3                     <pip>
google-auth               1.27.0                    <pip>
google-auth-oauthlib      0.4.2                     <pip>
google-pasta              0.2.0                     <pip>
grpcio                    1.35.0                    <pip>
h5py                      2.10.0                    <pip>
icu                       68.1                 h0e60522_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
idna                      2.10                      <pip>
importlib-metadata        3.7.0                     <pip>
intel-openmp              2020.3             h57928b3_311    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
jasper                    2.0.14               h77af90b_2    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
jpeg                      9d                   h8ffe710_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
Keras-Preprocessing       1.1.2                     <pip>
kiwisolver                1.3.1            py37h8c56517_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
lcms2                     2.12                 h2a16943_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libblas                   3.9.0                     8_mkl    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libcblas                  3.9.0                     8_mkl    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libclang                  11.0.1          default_h5c34c98_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
liblapack                 3.9.0                     8_mkl    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
liblapacke                3.9.0                     8_mkl    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libopencv                 4.5.1            py37ha0199f4_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libpng                    1.6.37               h1d00b33_2    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libtiff                   4.2.0                hc10be44_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
libwebp-base              1.2.0                h8ffe710_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
lz4-c                     1.9.3                h8ffe710_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
Markdown                  3.3.4                     <pip>
matplotlib                3.3.4            py37h03978a9_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
matplotlib-base           3.3.4            py37h3379fd5_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
mkl                       2020.4             hb70f87d_311    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
numpy                     1.20.1           py37hd20adf4_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
oauthlib                  3.1.0                     <pip>
olefile                   0.46               pyh9f0ad1d_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
opencv                    4.5.1            py37h03978a9_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
openssl                   1.1.1j               h8ffe710_0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
opt-einsum                3.3.0                     <pip>
pillow                    8.1.0            py37h96663a1_2    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pip                       21.0.1             pyhd8ed1ab_0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
protobuf                  3.15.2                    <pip>
py-opencv                 4.5.1            py37heaed05f_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pyasn1                    0.4.8                     <pip>
pyasn1-modules            0.2.8                     <pip>
pyparsing                 2.4.7              pyh9f0ad1d_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pyqt                      5.12.3           py37h03978a9_7    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pyqt-impl                 5.12.3           py37hf2a7229_7    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pyqt5-sip                 4.19.18          py37hf2a7229_7    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pyqtchart                 5.12             py37hf2a7229_7    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
pyqtwebengine             5.12.1           py37hf2a7229_7    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
python                    3.7.9           h7840368_100_cpython    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
python-dateutil           2.8.1                      py_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
python_abi                3.7                     1_cp37m    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
qt                        5.12.9               h5909a2a_4    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
requests                  2.25.1                    <pip>
requests-oauthlib         1.3.0                     <pip>
rsa                       4.7.2                     <pip>
scipy                     1.4.1                     <pip>
setuptools                49.6.0           py37h03978a9_3    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
six                       1.15.0             pyh9f0ad1d_0    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
sqlite                    3.34.0               h8ffe710_0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
tensorboard               2.2.2                     <pip>
tensorboard-plugin-wit    1.8.0                     <pip>
tensorflow-gpu            2.2.0                     <pip>
tensorflow-gpu-estimator  2.2.0                     <pip>
termcolor                 1.1.0                     <pip>
tk                        8.6.10               h8ffe710_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
tornado                   6.1              py37hcc03f2d_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
typing-extensions         3.7.4.3                   <pip>
urllib3                   1.26.3                    <pip>
vc                        14.2                 hb210afc_3    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
vs2015_runtime            14.28.29325          h5e1d092_3    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
Werkzeug                  1.0.1                     <pip>
wheel                     0.36.2             pyhd3deb0d_0    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
wincertstore              0.2             py37h03978a9_1006    https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
wrapt                     1.12.1                    <pip>
xz                        5.2.5                h62dcd97_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
zipp                      3.4.0                     <pip>
zlib                      1.2.11            h62dcd97_1010    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge
zstd                      1.4.8                h4e2f164_1    https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge

0.用pycharm打开时，记得切换对应的虚拟环境

1.VOCdevkit

2.Main中产生4个.txt

3.voc_annotation.py文件，将classes修改为自己的类别，然后运行生成yolo格式的标注文件

4.删去2007_

5复制yolov3.weights到根目录下，并在终端activate 虚拟环境 运行

python convert.py yolov3.cfg yolov3.weights model_data/yolo.h5

6.修改model_data中的coco_classes和voc_classes

7.修改yolov3.cfg中三处yolo附件的filters=3*(5+类别数量)，和 classes

8.将之前生成的yolo.h5(存在于model_data中)改名为yolo_weights.h5，对应的settings.py中的DEFAULT_MODEL_PATH的值也会变成yolo_weights.h5

9.在根目录下新建logs，在logs下新建000文件夹

10.运行train.py

11.在logs文件夹下，找到trained_weights_final.h5，将其复制到model_data中

12.更改settings中DEFAULT_MODEL_PATH，让其等于trained_weights_final.h5

13.运行python yolo_video.py --image即可