14
14

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 5 years have passed since last update.

CenterNetのwindowsでの実行

Last updated at Posted at 2019-07-20

はじめに

 CornetNet-Liteの記事でCenterNetをやるといっていたのですが、その後、Semantic SegmentationのLEDNetと、BiSeNetをやっていたので、時間が空いてしまいました。しかし、自分の記憶のためにもCenterNetも(できれば、LEDNetもBiSeNetも)書いておくべきだと思い、書いておきます。

#1.CenterNetとは
 Objects as Pointsの方です。CenterNet: Keypoint Triplets for Object Detectionではないです。1日の違いなので、名前が被ったのはお互い不幸でしたね。(私も混乱しました。)

 CenterNetの詳細は私が的外れな解説をするより、Slide ShaderにDL輪講会のものがありますので、そちらを参考に、原論文をあたってください。
 CornerNetもそうですが、従来のやり方で多数出てきていた、アンカーボックスとNMSを使わずに、一番それらしいボックスをいかにして出すかということが肝なんですね。

 では、githubからクローニングして、推論していきましょう。

#2.インストール
 論文著者のgithubINSTALL.mdがあるので、それに沿ってやっていきましょう。
##2.0.仮想環境構築

(base) F:\Users\sounansu\Anaconda3>conda create --name CenterNet python=3.6
Collecting package metadata (current_repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: F:\Users\sounansu\Anaconda3\envs\CenterNet

  added / updated specs:
    - python=3.6


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    certifi-2019.6.16          |           py36_0         151 KB
    pip-19.1.1                 |           py36_0         1.6 MB
    python-3.6.8               |       h9f7ef89_7        15.9 MB
    setuptools-41.0.1          |           py36_0         521 KB
    sqlite-3.28.0              |       he774522_0         616 KB
    vc-14.1                    |       h0510ff6_4           6 KB
    vs2015_runtime-14.15.26706 |       h3a45250_4         1.1 MB
    wheel-0.33.4               |           py36_0          57 KB
    wincertstore-0.2           |   py36h7fe50ca_0          14 KB
    ------------------------------------------------------------
                                           Total:        20.0 MB

The following NEW packages will be INSTALLED:

  certifi            pkgs/main/win-64::certifi-2019.6.16-py36_0
  pip                pkgs/main/win-64::pip-19.1.1-py36_0
  python             pkgs/main/win-64::python-3.6.8-h9f7ef89_7
  setuptools         pkgs/main/win-64::setuptools-41.0.1-py36_0
  sqlite             pkgs/main/win-64::sqlite-3.28.0-he774522_0
  vc                 pkgs/main/win-64::vc-14.1-h0510ff6_4
  vs2015_runtime     pkgs/main/win-64::vs2015_runtime-14.15.26706-h3a45250_4
  wheel              pkgs/main/win-64::wheel-0.33.4-py36_0
  wincertstore       pkgs/main/win-64::wincertstore-0.2-py36h7fe50ca_0


Proceed ([y]/n)? y


Downloading and Extracting Packages
pip-19.1.1           | 1.6 MB    | ############################################################ | 100%
vs2015_runtime-14.15 | 1.1 MB    | ############################################################ | 100%
python-3.6.8         | 15.9 MB   | ############################################################ | 100%
sqlite-3.28.0        | 616 KB    | ############################################################ | 100%
wheel-0.33.4         | 57 KB     | ############################################################ | 100%
setuptools-41.0.1    | 521 KB    | ############################################################ | 100%
certifi-2019.6.16    | 151 KB    | ############################################################ | 100%
vc-14.1              | 6 KB      | ############################################################ | 100%
wincertstore-0.2     | 14 KB     | ############################################################ | 100%
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
#
# To activate this environment, use
#
#     $ conda activate CenterNet
#
# To deactivate an active environment, use
#
#     $ conda deactivate


(base) F:\Users\sounansu\Anaconda3>conda activate CenterNet

##2.1.Pytorch0.4.1インストール
 ここでPyTorch0.4.1を指定してインストールするのですが、、、。

(CenterNet) F:\Users\sounansu\Anaconda3>conda install pytorch=0.4.1 torchvision -c pytorch
略
done

(CenterNet) F:\Users\sounansu\Anaconda3>

このあと、
And disable cudnn batch normalizatio
のために

# PYTORCH=/path/to/pytorch # usually ~/anaconda3/envs/CenterNet/lib/python3.6/site-packages/
# for pytorch v0.4.1
sed -i "1254s/torch\.backends\.cudnn\.enabled/False/g" ${PYTORCH}/torch/nn/functional.py

と書かれていますが、私のWindowsの環境ではfunctional.pyは
F:\Users\sounansu\Anaconda3\envs\CenterNet\Lib\site-packages\torch\nn\functional.py
でした。行は1254行でした。(どうもこの辺りですでに、PyTorch0.4.1ではなく、PyTorch1.1になっていたようです。)

##2.2.COCOAPIのインストール
 ここで、COCOAPIのインストールですが、CenterNetのディレクトリに入れたいので、先にCenterNetをクローニングします。(git が入っていなかったので、conda install gitを先にしました。)
##2.3.クローニング

(CenterNet) F:\Users\sounansu\Anaconda3>git clone https://github.com/xingyizhou/CenterNet CenterNet
Cloning into 'CenterNet'...
remote: Enumerating objects: 311, done.
remote: Total 311 (delta 0), reused 0 (delta 0), pack-reused 311
MiB | 564.00 KiB/s
Receiving objects: 100% (311/311), 6.25 MiB | 625.00 KiB/s, done.
Resolving deltas: 100% (120/120), done.

(CenterNet) F:\Users\sounansu\Anaconda3>

 改めて、COCOAPIをインストール

(CenterNet) F:\Users\sounansu\Anaconda3>cd CenterNet

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet>git clone https://github.com/cocodataset/cocoapi.git
Cloning into 'cocoapi'...
remote: Enumerating objects: 953, done.
remote: Total 953 (delta 0), reused 0 (delta 0), pack-reused 953 eceiving objects: 100% (953/953), 10.97Receiving objects: 100% (953/953), 11.70 MiB | 1.74 MiB/s, done.

Resolving deltas: 100% (565/565), done.

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet>cd cocoapi\PythonAPI

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\cocoapi\PythonAPI>make
'make' は、内部コマンドまたは外部コマンド、
操作可能なプログラムまたはバッチ ファイルとして認識されていません。

 conda install をつかってmakeを conda install makeしようとすると、channel にないといわれるので、Makefileのなかをみて、

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\cocoapi\PythonAPI>type Makefile
all:
    # install pycocotools locally
        python setup.py build_ext --inplace
        rm -rf build

install:
        # install pycocotools to the Python site-packages
        python setup.py build_ext install
        rm -rf build

all:の方だから、

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\cocoapi\PythonAPI>python setup.py build_ext --inplace
running build_ext
cythoning pycocotools/_mask.pyx to pycocotools\_mask.c
f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\Cython\Compiler\Main.py:367: FutureWarning: Cython directive 'language_level' not set, using 2 for now (Py2). This will change in a later release! File: F:\Users\sounansu\Anaconda3\CenterNet\cocoapi\PythonAPI\pycocotools\_mask.pyx
  tree = Parsing.p_module(s, pxd, full_module_name)
building 'pycocotools._mask' extension
creating build
creating build\temp.win-amd64-3.6
creating build\temp.win-amd64-3.6\common
creating build\temp.win-amd64-3.6\Release
creating build\temp.win-amd64-3.6\Release\pycocotools
C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\HostX86\x64\cl.exe /c /nologo /Ox /W3 /GL /DNDEBUG /MD -If:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\numpy\core\include -I../common -IF:\Users\sounansu\Anaconda3\envs\CenterNet\include -IF:\Users\sounansu\Anaconda3\envs\CenterNet\include "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\ATLMFC\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\include" "-IC:\Program Files (x86)\Windows Kits\NETFXSDK\4.6.1\include\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\shared" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\winrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\cppwinrt" "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include\include" /Tc../common/maskApi.c /Fobuild\temp.win-amd64-3.6\Release\../common/maskApi.obj -Wno-cpp -Wno-unused-function -std=c99
cl : コマンド ライン error D8021 : 数値型引数 '/Wno-cpp' は無効です。
error: command 'C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\VC\\Tools\\MSVC\\14.16.27023\\bin\\HostX86\\x64\\cl.exe' failed with exit status 2

とすると、なんか顔なじみのエラー(`/Wno-cpp`は無効です)が出てきましたね。というより、CornerNet-LiteのWindows 10での学習(ようやくトレーニング) 1.MS COCO APIsのインストールそのものですから、

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\cocoapi\PythonAPI>git diff
diff --git a/PythonAPI/setup.py b/PythonAPI/setup.py
index dbf0093..cd37940 100644
--- a/PythonAPI/setup.py
+++ b/PythonAPI/setup.py
@@ -9,7 +9,7 @@ ext_modules = [
         'pycocotools._mask',
         sources=['../common/maskApi.c', 'pycocotools/_mask.pyx'],
         include_dirs = [np.get_include(), '../common'],
-        extra_compile_args=['-Wno-cpp', '-Wno-unused-function', '-std=c99'],
+        extra_compile_args={'gcc': ['/Qstd=c99']},
     )
 ]

とCornerNet-Liteのときと同じことをしました。
 改めて

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\cocoapi\PythonAPI>python setup.py build_ext install
略
Finished processing dependencies for pycocotools==2.0

 COCO APIは無事にインストールされました。

##2.4.requirementsのインストール
 INSTALL.mdと順序を逆にしたので次はrequirements.txtのインストールです。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet>pip install -r requirements.txt
Requirement already satisfied: opencv-python in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from -r requirements.txt (line 1)) (4.1.0.25)
Requirement already satisfied: Cython in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from -r requirements.txt (line 2)) (0.29.7)
Requirement already satisfied: numba in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from -r requirements.txt (line 3)) (0.43.1)
Requirement already satisfied: progress in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from -r requirements.txt (line 4)) (1.5)
Requirement already satisfied: matplotlib in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from -r requirements.txt (line 5)) (3.1.0)
Collecting easydict (from -r requirements.txt (line 6))
  Downloading https://files.pythonhosted.org/packages/4c/c5/5757886c4f538c1b3f95f6745499a24bffa389a805dee92d093e2d9ba7db/easydict-1.9.tar.gz
Collecting scipy (from -r requirements.txt (line 7))
  Using cached https://files.pythonhosted.org/packages/9e/fd/9a995b7fc18c6c17ce570b3cfdabffbd2718e4f1830e94777c4fd66e1179/scipy-1.3.0-cp36-cp36m-win_amd64.whl
Requirement already satisfied: numpy>=1.11.3 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from opencv-python->-r requirements.txt (line 1)) (1.16.3)
Requirement already satisfied: llvmlite>=0.28.0dev0 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from numba->-r requirements.txt (line 3)) (0.28.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from matplotlib->-r requirements.txt (line 5)) (2.4.0)
Requirement already satisfied: python-dateutil>=2.1 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from matplotlib->-r requirements.txt (line 5)) (2.8.0)
Requirement already satisfied: cycler>=0.10 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from matplotlib->-r requirements.txt (line 5)) (0.10.0)
Requirement already satisfied: kiwisolver>=1.0.1 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from matplotlib->-r requirements.txt (line 5)) (1.1.0)
Requirement already satisfied: six>=1.5 in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from python-dateutil>=2.1->matplotlib->-r requirements.txt (line 5)) (1.12.0)
Requirement already satisfied: setuptools in f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages (from kiwisolver>=1.0.1->matplotlib->-r requirements.txt (line 5)) (41.0.1)
Building wheels for collected packages: easydict
  Building wheel for easydict (setup.py) ... done
  Stored in directory: C:\Users\sounansu\AppData\Local\pip\Cache\wheels\9a\88\ec\085d92753646b0eda1b7df49c7afe51a6ecc496556d3012e2e
Successfully built easydict
Installing collected packages: easydict, scipy
Successfully installed easydict-1.9 scipy-1.3.0

##2.5.デフォーマブルコンボリューションのインストール
次はDCNv2のインストールなわけですが、make.shを実行すると、別のウインドウが開いてよくわからないので、make.shの中身を一つ一つ実行しました。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>type make.sh
#!/usr/bin/env bash
cd src/cuda

# compile dcn
nvcc -c -o dcn_v2_im2col_cuda.cu.o dcn_v2_im2col_cuda.cu -x cu -Xcompiler -fPIC
nvcc -c -o dcn_v2_im2col_cuda_double.cu.o dcn_v2_im2col_cuda_double.cu -x cu -Xcompiler -fPIC

# compile dcn-roi-pooling
nvcc -c -o dcn_v2_psroi_pooling_cuda.cu.o dcn_v2_psroi_pooling_cuda.cu -x cu -Xcompiler -fPIC
nvcc -c -o dcn_v2_psroi_pooling_cuda_double.cu.o dcn_v2_psroi_pooling_cuda_double.cu -x cu -Xcompiler -fPIC

cd -
python build.py
python build_double.py

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>cd src\cuda

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2\src\cuda>nvcc -c -o dcn_v2_im2col_cuda.cu.o dcn_v2_im2col_cuda.cu -x cu -Xcompiler -fPIC
略
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2\src\cuda>nvcc -c -o dcn_v2_im2col_cuda_double.cu.o dcn_v2_im2col_cuda_double.cu -x cu -Xcompiler -fPIC
略
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2\src\cuda>nvcc -c -o dcn_v2_psroi_pooling_cuda.cu.o dcn_v2_psroi_pooling_cuda.cu -x cu -Xcompiler -fPIC
略
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2\src\cuda>nvcc -c -o dcn_v2_psroi_pooling_cuda_double.cu.o dcn_v2_psroi_pooling_cuda_double.cu -x cu -Xcompiler -fPIC
略
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2\src\cuda>cd ..\..\
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>python build.py
Traceback (most recent call last):
  File "build.py", line 3, in <module>
    from torch.utils.ffi import create_extension
  File "f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\torch\utils\ffi\__init__.py", line 1, in <module>
    raise ImportError("torch.utils.ffi is deprecated. Please use cpp extensions instead.")
ImportError: torch.utils.ffi is deprecated. Please use cpp extensions instead.

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>

あらあら、まあまあ、なんかエラーが出ています。issueをみると、pytorchのバージョンが新しいとこのようなエラーが出てその場合はDCNv2の新しいのを取ってくるようにと書いています。pytorchはちゃん0.41を入れたつもりなのですが。。。。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>python
Python 3.6.8 |Anaconda, Inc.| (default, Feb 21 2019, 18:30:04) [MSC v.1916 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> print(torch.__version__)
1.1.0
>>>

いつの間に、、、、。
pytorchが1.1だと、issue7にあるように、DCNv2をご本尊からとってこないとダメ見たいです。(初めからissue7を紹介しておけば、私のこの記事はいらない。。。。)

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet>cd src\lib\models\networks
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks>mv DCNv2 DCNv2_old
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks>git clone https://github.com/CharlesShang/DCNv2
Cloning into 'DCNv2'...
remote: Enumerating objects: 177, done.
remote: Total 177 (delta 0), reused 0 (delta 0), pack-reused 177 eceiving objects:  77% (137/177), 1.24 MiB | 270.00 KiB/s
Receiving objects: 100% (177/177), 1.39 MiB | 278.00 KiB/s, done.
Resolving deltas: 100% (106/106), done.
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks>

ここで、issue7の通りファイルを修正

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks>cd DCNv2
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>git diff
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>git diff
diff --git a/src/cuda/dcn_v2_cuda.cu b/src/cuda/dcn_v2_cuda.cu
index 767ed8f..5a6f7f4 100644
--- a/src/cuda/dcn_v2_cuda.cu
+++ b/src/cuda/dcn_v2_cuda.cu
@@ -8,7 +8,8 @@
 #include <THC/THCAtomics.cuh>
 #include <THC/THCDeviceUtils.cuh>

-extern THCState *state;
+//extern THCState *state;
+THCState *state = at::globalContext().lazyInitCUDA();

 // author: Charles Shang
 // https://github.com/torch/cunn/blob/master/lib/THCUNN/generic/SpatialConvolutionMM.cu
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\models\networks\DCNv2>python setup.py build develop

ものすごい量のログが出ます。がうまくいったようです。

(2.6飛ばして)
##2.7.学習済み重みの取得
 学習済みのデータはMODEL_ZOOからリンクされていて、グーグルドライブに格納されているようなので、好きなのをとってきて、、、というより、
いちいちとるのが面倒なので、全部取ってきました。。。

これで、下準備は終わりです。
#3.さあ始めましょう。
 私はもうすでに、MSCOCOやPASCAL VOCのデータはダウンロードしていますが、もしダウンロードしていないのなら、Dataset preparationを参考にしてダウンロードしてください。
##3.1.Benchmark evaluation
Benchmark evaluationのCOCOをやってみましょう。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>cd
F:\Users\sounansu\Anaconda3\CenterNet\src

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python test.py ctdet --exp_id coco_dla --keep_res --load_model ..\models\ctdet_coco_dla_2x.pth
Traceback (most recent call last):
  File "test.py", line 15, in <module>
    from external.nms import soft_nms
ModuleNotFoundError: No module named 'external.nms'

 たしかnmsのコンパイルってオプションって書いてたのですが。。。
issue71. build nmsを参考にコンパイルしましょう。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>cd lib\external

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\external>python setup.py build_ext --inplace
Compiling nms.pyx because it changed.
[1/1] Cythonizing nms.pyx
f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\Cython\Compiler\Main.py:367: FutureWarning: Cython directive 'language_level' not set, using 2 for now (Py2). This will change in a later release! File: F:\Users\sounansu\Anaconda3\CenterNet\src\lib\external\nms.pyx
  tree = Parsing.p_module(s, pxd, full_module_name)
running build_ext
building 'nms' extension
creating build
creating build\temp.win-amd64-3.6
creating build\temp.win-amd64-3.6\Release
C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\HostX86\x64\cl.exe /c /nologo /Ox /W3 /GL /DNDEBUG /MD -If:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\numpy\core\include -IF:\Users\sounansu\Anaconda3\envs\CenterNet\include -IF:\Users\sounansu\Anaconda3\envs\CenterNet\include "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\ATLMFC\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\include" "-IC:\Program Files (x86)\Windows Kits\NETFXSDK\4.6.1\include\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\shared" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\winrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\cppwinrt" "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include\include" /Tcnms.c /Fobuild\temp.win-amd64-3.6\Release\nms.obj -Wno-cpp -Wno-unused-function
cl : コマンド ライン error D8021 : 数値型引数 '/Wno-cpp' は無効です。
error: command 'C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\VC\\Tools\\MSVC\\14.16.27023\\bin\\HostX86\\x64\\cl.exe' failed with exit status 2

まぁ、windowsでやってればよく出るエラーです。issue7にあるようにextra_compile_argsをコメントアウトしましょう。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\external>git diff setup.py
diff --git a/src/lib/external/setup.py b/src/lib/external/setup.py
index c4d2571..3b140cb 100644
--- a/src/lib/external/setup.py
+++ b/src/lib/external/setup.py
@@ -7,7 +7,7 @@ extensions = [
     Extension(
         "nms",
         ["nms.pyx"],
-        extra_compile_args=["-Wno-cpp", "-Wno-unused-function"]
+#        extra_compile_args=["-Wno-cpp", "-Wno-unused-function"]
     )
 ]
(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src\lib\external>python setup.py build_ext --inplace
running build_ext

(成功したので略)

   ライブラリ build\temp.win-amd64-3.6\Release\nms.cp36-win_amd64.lib とオブジェクト build\temp.win-amd64-3.6\Release\nms.cp36-win_amd64.exp を作成中
コード生成しています。
コード生成が終了しました。

 さて、evaluationをやってみましょう。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python test.py ctdet --exp_id coco_dla --keep_res --load_model ../models/ctdet_coco_dla_2x.pth

(エラーがたくさん)

あまりにもエラーが多すぎるので、issue7にあるdemo.pyの方をやってみると、

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python demo.py ctdet --demo ../images/17790319373_bd19b24cfc_k.jpg --load_model ../models/ctdet_coco_dla_2x.pth --debug 2
Fix size testing.
training chunk_sizes: [1]
The output will be saved to  F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\exp\ctdet\default
heads {'hm': 80, 'wh': 2, 'reg': 2}
Creating model...
loaded ../models/ctdet_coco_dla_2x.pth, epoch 230
tot 2.330s |load 0.000s |pre 0.031s |net 2.217s |dec 0.004s |post 0.078s |merge 0.000s |

画像が表示されました。output.png

evaluationができないので途方に暮れて、Qiitaにも投稿できずにいたのですが、一応issue126でへたっぴな英語で質問したところ、
--not_prefetch_testをつけろと言われました。
 issue126ではこれでできるように書いていますが、このままではエラーがでます。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python test.py ctdet --not_prefetch_test --exp_id coco_dla --keep_res --load_model ..\models\ctdet_coco_dla_2x.pth
Keep resolution testing.
training chunk_sizes: [32]
The output will be saved to  F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\exp\ctdet\coco_dla
heads {'hm': 80, 'wh': 2, 'reg': 2}
Namespace(K=100, aggr_weight=0.0, agnostic_ex=False, arch='dla_34', aug_ddd=0.5, aug_rot=0, batch_size=32, cat_spec_wh=False, center_thresh=0.1, chunk_sizes=[32], data_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\data', dataset='coco', debug=0, debug_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla\\debug', debugger_theme='white', demo='', dense_hp=False, dense_wh=False, dep_weight=1, dim_weight=1, down_ratio=4, eval_oracle_dep=False, eval_oracle_hm=False, eval_oracle_hmhp=False, eval_oracle_hp_offset=False, eval_oracle_kps=False, eval_oracle_offset=False, eval_oracle_wh=False, exp_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet', exp_id='coco_dla', fix_res=False, flip=0.5, flip_test=False, gpus=[0], gpus_str='0', head_conv=256, heads={'hm': 80, 'wh': 2, 'reg': 2}, hide_data_time=False, hm_hp=True, hm_hp_weight=1, hm_weight=1, hp_weight=1, input_h=512, input_res=512, input_w=512, keep_res=True, kitti_split='3dop', load_model='..\\models\\ctdet_coco_dla_2x.pth', lr=0.000125, lr_step=[90, 120], master_batch_size=32, mean=array([[[0.40789655, 0.44719303, 0.47026116]]], dtype=float32), metric='loss', mse_loss=False, nms=False, no_color_aug=False, norm_wh=False, not_cuda_benchmark=False, not_hm_hp=False, not_prefetch_test=True, not_rand_crop=False, not_reg_bbox=False, not_reg_hp_offset=False, not_reg_offset=False, num_classes=80, num_epochs=140, num_iters=-1, num_stacks=1, num_workers=4, off_weight=1, output_h=128, output_res=128, output_w=128, pad=31, peak_thresh=0.2, print_iter=0, rect_mask=False, reg_bbox=True, reg_hp_offset=True, reg_loss='l1', reg_offset=True, resume=False, root_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..', rot_weight=1, rotate=0, save_all=False, save_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla', scale=0.4, scores_thresh=0.1, seed=317, shift=0.1, std=array([[[0.2886383 , 0.27408165, 0.27809834]]], dtype=float32), task='ctdet', test=False, test_scales=[1.0], trainval=False, val_intervals=5, vis_thresh=0.3, wh_weight=0.1)
==> initializing coco 2017 val data.
loading annotations into memory...
Done (t=0.71s)
creating index...
index created!
Loaded val 5000 samples
Creating model...
loaded ..\models\ctdet_coco_dla_2x.pth, epoch 230
coco_dlaTraceback (most recent call last):
  File "test.py", line 124, in <module>
    test(opt)
  File "test.py", line 108, in test
    ret = detector.run(img_path)
  File "F:\Users\sounansu\Anaconda3\CenterNet\src\lib\detectors\base_detector.py", line 105, in run
    images, meta = self.pre_process(image, scale, meta)
  File "F:\Users\sounansu\Anaconda3\CenterNet\src\lib\detectors\base_detector.py", line 38, in pre_process
    height, width = image.shape[0:2]
AttributeError: 'NoneType' object has no attribute 'shape'

なぜshapeがない。。。と、過去にこの手のエラーがあった場合は、パスやファイル名を間違った時なので、src\lib\datasets\dataset\coco.pydata_dirimg_dirprint()文で調べると

F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco\val2017

あれ、Dataset preparationと違う。。。(これがwindowsのせいかよくわかりません。)coco.pyをDataset preparationに会うように変更して、

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>git diff lib\datasets\dataset\coco.py
diff --git a/src/lib/datasets/dataset/coco.py b/src/lib/datasets/dataset/coco.py
index d0efc53..a89628c 100644
--- a/src/lib/datasets/dataset/coco.py
+++ b/src/lib/datasets/dataset/coco.py
@@ -21,7 +21,10 @@ class COCO(data.Dataset):
   def __init__(self, opt, split):
     super(COCO, self).__init__()
     self.data_dir = os.path.join(opt.data_dir, 'coco')
-    self.img_dir = os.path.join(self.data_dir, '{}2017'.format(split))
+    self.img_dir = os.path.join(self.data_dir, 'images')
+    self.img_dir = os.path.join(self.img_dir, '{}2017'.format(split))
     if split == 'test':
       self.annot_path = os.path.join(
           self.data_dir, 'annotations',


(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python test.py ctdet --not_prefetch_test --exp_id coco_dla --keep_res --load_model ..\models\ctdet_coco_dla_2x.pth
Keep resolution testing.
training chunk_sizes: [32]
The output will be saved to  F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\exp\ctdet\coco_dla
heads {'hm': 80, 'wh': 2, 'reg': 2}
Namespace(K=100, aggr_weight=0.0, agnostic_ex=False, arch='dla_34', aug_ddd=0.5, aug_rot=0, batch_size=32, cat_spec_wh=False, center_thresh=0.1, chunk_sizes=[32], data_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\data', dataset='coco', debug=0, debug_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla\\debug', debugger_theme='white', demo='', dense_hp=False, dense_wh=False, dep_weight=1, dim_weight=1, down_ratio=4, eval_oracle_dep=False, eval_oracle_hm=False, eval_oracle_hmhp=False, eval_oracle_hp_offset=False, eval_oracle_kps=False, eval_oracle_offset=False, eval_oracle_wh=False, exp_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet', exp_id='coco_dla', fix_res=False, flip=0.5, flip_test=False, gpus=[0], gpus_str='0', head_conv=256, heads={'hm': 80, 'wh': 2, 'reg': 2}, hide_data_time=False, hm_hp=True, hm_hp_weight=1, hm_weight=1, hp_weight=1, input_h=512, input_res=512, input_w=512, keep_res=True, kitti_split='3dop', load_model='..\\models\\ctdet_coco_dla_2x.pth', lr=0.000125, lr_step=[90, 120], master_batch_size=32, mean=array([[[0.40789655, 0.44719303, 0.47026116]]], dtype=float32), metric='loss', mse_loss=False, nms=False, no_color_aug=False, norm_wh=False, not_cuda_benchmark=False, not_hm_hp=False, not_prefetch_test=True, not_rand_crop=False, not_reg_bbox=False, not_reg_hp_offset=False, not_reg_offset=False, num_classes=80, num_epochs=140, num_iters=-1, num_stacks=1, num_workers=4, off_weight=1, output_h=128, output_res=128, output_w=128, pad=31, peak_thresh=0.2, print_iter=0, rect_mask=False, reg_bbox=True, reg_hp_offset=True, reg_loss='l1', reg_offset=True, resume=False, root_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..', rot_weight=1, rotate=0, save_all=False, save_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla', scale=0.4, scores_thresh=0.1, seed=317, shift=0.1, std=array([[[0.2886383 , 0.27408165, 0.27809834]]], dtype=float32), task='ctdet', test=False, test_scales=[1.0], trainval=False, val_intervals=5, vis_thresh=0.3, wh_weight=0.1)
==> initializing coco 2017 val data.
loading annotations into memory...
Done (t=0.70s)
creating index...
index created!
Loaded val 5000 samples
Creating model...
loaded ..\models\ctdet_coco_dla_2x.pth, epoch 230
coco_dla |################################| [4999/5000]|Tot: 0:08:14 |ETA: 0:00:01 |tot 0.098 |load 0.022 |pre 0.025 |net 0.045 |dec 0.004 |post 0.003 |merge 0.000
Loading and preparing results...
DONE (t=2.97s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type *bbox*
DONE (t=54.50s).
Accumulating evaluation results...
DONE (t=11.94s).
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.374
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.551
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.408
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.206
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.420
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.506
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.317
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.521
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.551
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.336
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.594
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.737

MODEL_ZOOにある、37.4%に一致しました。
##3.2.トレーニング
Trainingにあるように、でもGPUは一個なので以下のように変更して実行しました。

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python main.py ctdet --exp_id coco_dla --batch_size 32 --master_batch 15 --lr 1.25e-4  --gpus  0
Fix size testing.
training chunk_sizes: [15]
The output will be saved to  F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\exp\ctdet\coco_dla
heads {'hm': 80, 'wh': 2, 'reg': 2}
Namespace(K=100, aggr_weight=0.0, agnostic_ex=False, arch='dla_34', aug_ddd=0.5, aug_rot=0, batch_size=32, cat_spec_wh=False, center_thresh=0.1, chunk_sizes=[15], data_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\data', dataset='coco', debug=0, debug_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla\\debug', debugger_theme='white', demo='', dense_hp=False, dense_wh=False, dep_weight=1, dim_weight=1, down_ratio=4, eval_oracle_dep=False, eval_oracle_hm=False, eval_oracle_hmhp=False, eval_oracle_hp_offset=False, eval_oracle_kps=False, eval_oracle_offset=False, eval_oracle_wh=False, exp_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet', exp_id='coco_dla', fix_res=True, flip=0.5, flip_test=False, gpus=[0], gpus_str='0', head_conv=256, heads={'hm': 80, 'wh': 2, 'reg': 2}, hide_data_time=False, hm_hp=True, hm_hp_weight=1, hm_weight=1, hp_weight=1, input_h=512, input_res=512, input_w=512, keep_res=False, kitti_split='3dop', load_model='', lr=0.000125, lr_step=[90, 120], master_batch_size=15, mean=array([[[0.40789655, 0.44719303, 0.47026116]]], dtype=float32), metric='loss', mse_loss=False, nms=False, no_color_aug=False, norm_wh=False, not_cuda_benchmark=False, not_hm_hp=False, not_prefetch_test=False, not_rand_crop=False, not_reg_bbox=False, not_reg_hp_offset=False, not_reg_offset=False, num_classes=80, num_epochs=140, num_iters=-1, num_stacks=1, num_workers=4, off_weight=1, output_h=128, output_res=128, output_w=128, pad=31, peak_thresh=0.2, print_iter=0, rect_mask=False, reg_bbox=True, reg_hp_offset=True, reg_loss='l1', reg_offset=True, resume=False, root_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..', rot_weight=1, rotate=0, save_all=False, save_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla', scale=0.4, scores_thresh=0.1, seed=317, shift=0.1, std=array([[[0.2886383 , 0.27408165, 0.27809834]]], dtype=float32), task='ctdet', test=False, test_scales=[1.0], trainval=False, val_intervals=5, vis_thresh=0.3, wh_weight=0.1)
Creating model...
Setting up data...
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco\images\val2017
==> initializing coco 2017 val data.
loading annotations into memory...
Done (t=0.68s)
creating index...
index created!
Loaded val 5000 samples
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco\images\train2017
==> initializing coco 2017 train data.
loading annotations into memory...
Done (t=17.81s)
creating index...
index created!
Loaded train 118287 samples
Starting training...
ctdet/coco_dlaTraceback (most recent call last):
  File "main.py", line 102, in <module>
    main(opt)
  File "main.py", line 70, in main
    log_dict_train, _ = trainer.train(epoch, train_loader)
  File "F:\Users\sounansu\Anaconda3\CenterNet\src\lib\trains\base_trainer.py", line 119, in train
    return self.run_epoch('train', epoch, data_loader)
  File "F:\Users\sounansu\Anaconda3\CenterNet\src\lib\trains\base_trainer.py", line 61, in run_epoch
    for iter_id, batch in enumerate(data_loader):
  File "f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\torch\utils\data\dataloader.py", line 193, in __iter__
    return _DataLoaderIter(self)
  File "f:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\torch\utils\data\dataloader.py", line 469, in __init__
    w.start()
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\process.py", line 105, in start
    self._popen = self._Popen(self)
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\context.py", line 223, in _Popen
    return _default_context.get_context().Process._Popen(process_obj)
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\context.py", line 322, in _Popen
    return Popen(process_obj)
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\popen_spawn_win32.py", line 65, in __init__
    reduction.dump(process_obj, to_child)
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\reduction.py", line 60, in dump
    ForkingPickler(file, protocol).dump(obj)
AttributeError: Can't pickle local object 'get_dataset.<locals>.Dataset'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\spawn.py", line 105, in spawn_main
    exitcode = _main(fd)
  File "F:\Users\sounansu\Anaconda3\envs\CenterNet\lib\multiprocessing\spawn.py", line 115, in _main
    self = reduction.pickle.load(from_parent)
EOFError: Ran out of input

これも、issue126で質問した結果と、自分のRTX2070のメモリ容量に合わせて、

(CenterNet) F:\Users\sounansu\Anaconda3\CenterNet\src>python main.py ctdet --exp_id coco_dla --batch_size 11 --master_batch 11 --lr 1.25e-4 --gpus 0 --num_workers 0
Fix size testing.
training chunk_sizes: [11]
The output will be saved to  F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\exp\ctdet\coco_dla
heads {'hm': 80, 'wh': 2, 'reg': 2}
Namespace(K=100, aggr_weight=0.0, agnostic_ex=False, arch='dla_34', aug_ddd=0.5, aug_rot=0, batch_size=11, cat_spec_wh=False, center_thresh=0.1, chunk_sizes=[11], data_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\data', dataset='coco', debug=0, debug_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla\\debug', debugger_theme='white', demo='', dense_hp=False, dense_wh=False, dep_weight=1, dim_weight=1, down_ratio=4, eval_oracle_dep=False, eval_oracle_hm=False, eval_oracle_hmhp=False, eval_oracle_hp_offset=False, eval_oracle_kps=False, eval_oracle_offset=False, eval_oracle_wh=False, exp_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet', exp_id='coco_dla', fix_res=True, flip=0.5, flip_test=False, gpus=[0], gpus_str='0', head_conv=256, heads={'hm': 80, 'wh': 2, 'reg': 2}, hide_data_time=False, hm_hp=True, hm_hp_weight=1, hm_weight=1, hp_weight=1, input_h=512, input_res=512, input_w=512, keep_res=False, kitti_split='3dop', load_model='', lr=0.000125, lr_step=[90, 120], master_batch_size=11, mean=array([[[0.40789655, 0.44719303, 0.47026116]]], dtype=float32), metric='loss', mse_loss=False, nms=False, no_color_aug=False, norm_wh=False, not_cuda_benchmark=False, not_hm_hp=False, not_prefetch_test=False, not_rand_crop=False, not_reg_bbox=False, not_reg_hp_offset=False, not_reg_offset=False, num_classes=80, num_epochs=140, num_iters=-1, num_stacks=1, num_workers=0, off_weight=1, output_h=128, output_res=128, output_w=128, pad=31, peak_thresh=0.2, print_iter=0, rect_mask=False, reg_bbox=True, reg_hp_offset=True, reg_loss='l1', reg_offset=True, resume=False, root_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..', rot_weight=1, rotate=0, save_all=False, save_dir='F:\\Users\\sounansu\\Anaconda3\\CenterNet\\src\\lib\\..\\..\\exp\\ctdet\\coco_dla', scale=0.4, scores_thresh=0.1, seed=317, shift=0.1, std=array([[[0.2886383 , 0.27408165, 0.27809834]]], dtype=float32), task='ctdet', test=False, test_scales=[1.0], trainval=False, val_intervals=5, vis_thresh=0.3, wh_weight=0.1)
Creating model...
Setting up data...
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco\images\val2017
==> initializing coco 2017 val data.
loading annotations into memory...
Done (t=0.66s)
creating index...
index created!
Loaded val 5000 samples
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco
F:\Users\sounansu\Anaconda3\CenterNet\src\lib\..\..\data\coco\images\train2017
==> initializing coco 2017 train data.
loading annotations into memory...
Done (t=17.87s)
creating index...
index created!
Loaded train 118287 samples
Starting training...
ctdet/coco_dlaf:\users\sounansu\anaconda3\envs\centernet\lib\site-packages\torch\nn\_reduction.py:46: UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead.
  warnings.warn(warning.format(ret))
ctdet/coco_dla |                                | train: [1][3/10753]|Tot: 0:00:15 |ETA: 13:28:07 |loss 245.5500 |hm_loss 242.5728 |wh_loss 25.6510 |off_loss 0.4120 |Data 0.412s(0.408s) |Net 3.776s

無事トレーニングが始まりました。
#4.終わりに
 正直、途中であきらめていたのですが、どうにかトレーニングまでこぎつけました。CenterNetのいいところは、やはり、pose estimationもできることでしょうか。要はobject detectionではクラスのhot spotとオフセットとボックスのところのクラスを、関節にとかに割り当てたという感じでしょうか?なんか不思議です。
 Object DetectionはCenterNetでいったん休止して、つぎは最近のSemantic Segmentationをやっていきたい(LEDNetと、BiSeNetを実際はやっています。)です。SegNetとかをやっていたので、新しいものはスピードが驚異的にしか見えません。

では。

14
14
2

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
14
14

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?