當前最好的實例分割網絡非PANet莫屬，可是由於模型太新，網上的資料太少，最近的項目需要 實例分割，只能自己踩踩坑了，目前我還沒看到一篇關於PANet訓練的博客，只有幾篇講論文的。
環境：ubuntu16.04 anaconda3 pytorch0.4.0
將代碼down下來后，我在anaconda中新開了一個環境，用的是python3.6，如果用python2的話，在lib文件夾下的make.sh中的python3需要修改一下
然后一開始的時候我沒有進行make，直接進行測試，運行
python tools/test_net.py --dataset coco2017 --cfg configs/panet/e2e_panet_R-50-FPN_2x_mask.yaml --load_ckpt {path/to/your/checkpoint}
然后，報錯如下
Traceback (most recent call last):
File "tools/test_net.py", line 14, in <module>
from core.test_engine import run_inference
File "/home/wzy/Documents/PANet-master/lib/core/test_engine.py", line 39, in <module>
from modeling import model_builder
File "/home/wzy/Documents/PANet-master/lib/modeling/model_builder.py", line 11, in <module>
from model.roi_pooling.functions.roi_pool import RoIPoolFunction
File "/home/wzy/Documents/PANet-master/lib/model/roi_pooling/functions/roi_pool.py", line 3, in <module>
from .._ext import roi_pooling
File "/home/wzy/Documents/PANet-master/lib/model/roi_pooling/_ext/roi_pooling/__init__.py", line 3, in <module>
from ._roi_pooling import lib as _lib, ffi as _ffi
ImportError: /home/wzy/Documents/PANet-master/lib/model/roi_pooling/_ext/roi_pooling/_roi_pooling.so: undefined symbol: __cudaPopCallConfiguration
還是需要make一下的，結果這個環節更讓人頭大，一直報錯
running build_ext
building 'utils.cython_bbox' extension
creating build
creating build/temp.linux-x86_64-3.6
creating build/temp.linux-x86_64-3.6/utils
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/numpy/core/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c utils/cython_bbox.c -o build/temp.linux-x86_64-3.6/utils/cython_bbox.o -Wno-cpp
creating build/lib.linux-x86_64-3.6
creating build/lib.linux-x86_64-3.6/utils
gcc -pthread -shared -L/home/ustc/anaconda3/envs/py36/lib -Wl,-rpath=/home/ustc/anaconda3/envs/py36/lib,--no-as-needed -std=c99 build/temp.linux-x86_64-3.6/utils/cython_bbox.o -L/home/ustc/anaconda3/envs/py36/lib -lpython3.6m -o build/lib.linux-x86_64-3.6/utils/cython_bbox.cpython-36m-x86_64-linux-gnu.so
building 'utils.cython_nms' extension
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/numpy/core/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c utils/cython_nms.c -o build/temp.linux-x86_64-3.6/utils/cython_nms.o -Wno-cpp
gcc -pthread -shared -L/home/ustc/anaconda3/envs/py36/lib -Wl,-rpath=/home/ustc/anaconda3/envs/py36/lib,--no-as-needed -std=c99 build/temp.linux-x86_64-3.6/utils/cython_nms.o -L/home/ustc/anaconda3/envs/py36/lib -lpython3.6m -o build/lib.linux-x86_64-3.6/utils/cython_nms.cpython-36m-x86_64-linux-gnu.so
copying build/lib.linux-x86_64-3.6/utils/cython_bbox.cpython-36m-x86_64-linux-gnu.so -> utils
copying build/lib.linux-x86_64-3.6/utils/cython_nms.cpython-36m-x86_64-linux-gnu.so -> utils
Compiling nms kernels by nvcc...
Including CUDA code.
/home/ustc/文檔/PANet-master/lib/model/nms
['/home/ustc/文檔/PANet-master/lib/model/nms/src/nms_cuda_kernel.cu.o']
generating /tmp/tmpwpj4u4zb/_nms.c
running build_ext
building '_nms' extension
creating home
creating home/ustc
creating home/ustc/文檔
creating home/ustc/文檔/PANet-master
creating home/ustc/文檔/PANet-master/lib
creating home/ustc/文檔/PANet-master/lib/model
creating home/ustc/文檔/PANet-master/lib/model/nms
creating home/ustc/文檔/PANet-master/lib/model/nms/src
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c _nms.c -o ./_nms.o -std=c99
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c /home/ustc/文檔/PANet-master/lib/model/nms/src/nms_cuda.c -o ./home/ustc/文檔/PANet-master/lib/model/nms/src/nms_cuda.o -std=c99
/home/ustc/文檔/PANet-master/lib/model/nms/src/nms_cuda.c: In function ‘nms_cuda’:
/home/ustc/文檔/PANet-master/lib/model/nms/src/nms_cuda.c:14:22: error: dereferencing pointer to incomplete type ‘THTensor {aka struct THTensor}’
boxes_host->size[0],
^
Traceback (most recent call last):
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/unixccompiler.py", line 118, in _compile
extra_postargs)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/ccompiler.py", line 909, in spawn
spawn(cmd, dry_run=self.dry_run)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/spawn.py", line 36, in spawn
_spawn_posix(cmd, search_path, dry_run=dry_run)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/spawn.py", line 159, in _spawn_posix
% (cmd, exit_status))
distutils.errors.DistutilsExecError: command 'gcc' failed with exit status 1
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/ffiplatform.py", line 49, in _build
dist.run_command('build_ext')
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/dist.py", line 974, in run_command
cmd_obj.run()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 339, in run
self.build_extensions()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 448, in build_extensions
self._build_extensions_serial()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 473, in _build_extensions_serial
self.build_extension(ext)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 533, in build_extension
depends=ext.depends)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/ccompiler.py", line 574, in compile
self._compile(obj, src, ext, cc_args, extra_postargs, pp_opts)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/unixccompiler.py", line 120, in _compile
raise CompileError(msg)
distutils.errors.CompileError: command 'gcc' failed with exit status 1
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "build.py", line 37, in <module>
ffi.build()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/__init__.py", line 189, in build
_build_extension(ffi, cffi_wrapper_name, target_dir, verbose)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/__init__.py", line 111, in _build_extension
outfile = ffi.compile(tmpdir=tmpdir, verbose=verbose, target=libname)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/api.py", line 684, in compile
compiler_verbose=verbose, debug=debug, **kwds)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/recompiler.py", line 1484, in recompile
compiler_verbose, debug)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/ffiplatform.py", line 20, in compile
outputfilename = _build(tmpdir, ext, compiler_verbose, debug)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/ffiplatform.py", line 56, in _build
raise VerificationError('%s: %s' % (e.__class__.__name__, e))
cffi.error.VerificationError: CompileError: command 'gcc' failed with exit status 1
Compiling roi pooling kernels by nvcc...
Including CUDA code.
/home/ustc/文檔/PANet-master/lib/model/roi_pooling
generating /tmp/tmpy7_eurcc/_roi_pooling.c
running build_ext
building '_roi_pooling' extension
creating home
creating home/ustc
creating home/ustc/文檔
creating home/ustc/文檔/PANet-master
creating home/ustc/文檔/PANet-master/lib
creating home/ustc/文檔/PANet-master/lib/model
creating home/ustc/文檔/PANet-master/lib/model/roi_pooling
creating home/ustc/文檔/PANet-master/lib/model/roi_pooling/src
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c _roi_pooling.c -o ./_roi_pooling.o -std=c99
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c /home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling.c -o ./home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling.o -std=c99
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c /home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling_cuda.c -o ./home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling_cuda.o -std=c99
gcc -pthread -shared -L/home/ustc/anaconda3/envs/py36/lib -Wl,-rpath=/home/ustc/anaconda3/envs/py36/lib,--no-as-needed -std=c99 ./_roi_pooling.o ./home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling.o ./home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling_cuda.o /home/ustc/文檔/PANet-master/lib/model/roi_pooling/src/roi_pooling.cu.o -L/home/ustc/anaconda3/envs/py36/lib -lpython3.6m -o ./_roi_pooling.so
Compiling roi crop kernels by nvcc...
Including CUDA code.
/home/ustc/文檔/PANet-master/lib/model/roi_crop
generating /tmp/tmptm4aw2z1/_roi_crop.c
running build_ext
building '_roi_crop' extension
creating home
creating home/ustc
creating home/ustc/文檔
creating home/ustc/文檔/PANet-master
creating home/ustc/文檔/PANet-master/lib
creating home/ustc/文檔/PANet-master/lib/model
creating home/ustc/文檔/PANet-master/lib/model/roi_crop
creating home/ustc/文檔/PANet-master/lib/model/roi_crop/src
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c _roi_crop.c -o ./_roi_crop.o -std=c99
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c /home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c -o ./home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.o -std=c99
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c: In function ‘BilinearSamplerBHWD_updateOutput’:
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:10:30: error: dereferencing pointer to incomplete type ‘THTensor {aka struct THTensor}’
int batchsize = inputImages->size[0];
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c: In function ‘BilinearSamplerBHWD_updateGradInput’:
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:190:14: warning: unused variable ‘inBottomRight’ [-Wunused-variable]
real inBottomRight=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:189:14: warning: unused variable ‘inBottomLeft’ [-Wunused-variable]
real inBottomLeft=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:188:14: warning: unused variable ‘inTopRight’ [-Wunused-variable]
real inTopRight=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:187:14: warning: unused variable ‘inTopLeft’ [-Wunused-variable]
real inTopLeft=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:186:14: warning: unused variable ‘v’ [-Wunused-variable]
real v=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c: In function ‘BilinearSamplerBCHW_updateGradInput’:
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:441:14: warning: unused variable ‘inBottomRight’ [-Wunused-variable]
real inBottomRight=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:440:14: warning: unused variable ‘inBottomLeft’ [-Wunused-variable]
real inBottomLeft=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:439:14: warning: unused variable ‘inTopRight’ [-Wunused-variable]
real inTopRight=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:438:14: warning: unused variable ‘inTopLeft’ [-Wunused-variable]
real inTopLeft=0;
^
/home/ustc/文檔/PANet-master/lib/model/roi_crop/src/roi_crop.c:437:14: warning: unused variable ‘v’ [-Wunused-variable]
real v=0;
^
Traceback (most recent call last):
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/unixccompiler.py", line 118, in _compile
extra_postargs)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/ccompiler.py", line 909, in spawn
spawn(cmd, dry_run=self.dry_run)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/spawn.py", line 36, in spawn
_spawn_posix(cmd, search_path, dry_run=dry_run)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/spawn.py", line 159, in _spawn_posix
% (cmd, exit_status))
distutils.errors.DistutilsExecError: command 'gcc' failed with exit status 1
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/ffiplatform.py", line 49, in _build
dist.run_command('build_ext')
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/dist.py", line 974, in run_command
cmd_obj.run()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 339, in run
self.build_extensions()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 448, in build_extensions
self._build_extensions_serial()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 473, in _build_extensions_serial
self.build_extension(ext)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/command/build_ext.py", line 533, in build_extension
depends=ext.depends)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/ccompiler.py", line 574, in compile
self._compile(obj, src, ext, cc_args, extra_postargs, pp_opts)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/distutils/unixccompiler.py", line 120, in _compile
raise CompileError(msg)
distutils.errors.CompileError: command 'gcc' failed with exit status 1
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "build.py", line 36, in <module>
ffi.build()
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/__init__.py", line 189, in build
_build_extension(ffi, cffi_wrapper_name, target_dir, verbose)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/__init__.py", line 111, in _build_extension
outfile = ffi.compile(tmpdir=tmpdir, verbose=verbose, target=libname)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/api.py", line 684, in compile
compiler_verbose=verbose, debug=debug, **kwds)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/recompiler.py", line 1484, in recompile
compiler_verbose, debug)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/ffiplatform.py", line 20, in compile
outputfilename = _build(tmpdir, ext, compiler_verbose, debug)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/cffi/ffiplatform.py", line 56, in _build
raise VerificationError('%s: %s' % (e.__class__.__name__, e))
cffi.error.VerificationError: CompileError: command 'gcc' failed with exit status 1
Compiling roi align kernels by nvcc...
Including CUDA code.
/home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align
generating /tmp/tmperu9f6wr/_roi_align.c
running build_ext
building '_roi_align' extension
creating home
creating home/ustc
creating home/ustc/文檔
creating home/ustc/文檔/PANet-master
creating home/ustc/文檔/PANet-master/lib
creating home/ustc/文檔/PANet-master/lib/modeling
creating home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom
creating home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align
creating home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align/src
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c _roi_align.c -o ./_roi_align.o -std=c99
gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -std=c99 -fPIC -DWITH_CUDA -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/TH -I/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/utils/ffi/../../lib/include/THC -I/usr/local/cuda/include -I/home/ustc/anaconda3/envs/py36/include/python3.6m -c /home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align/src/roi_align_cuda.c -o ./home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align/src/roi_align_cuda.o -std=c99
gcc -pthread -shared -L/home/ustc/anaconda3/envs/py36/lib -Wl,-rpath=/home/ustc/anaconda3/envs/py36/lib,--no-as-needed -std=c99 ./_roi_align.o ./home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align/src/roi_align_cuda.o /home/ustc/文檔/PANet-master/lib/modeling/roi_xfrom/roi_align/src/roi_align_kernel.cu.o -L/home/ustc/anaconda3/envs/py36/lib -lpython3.6m -o ./_roi_align.so
這里我主要google了一下 cffi.error.VerificationError: CompileError: command 'gcc' failed with exit status 1
我注意到這是一個很多人都遇到過的問題，在faster-rcnn的pytorch實現中也有這個問題，似乎這個代碼的實現部分與faster-rcnn的有相同部分，然后我主要采用了兩步解決辦法，一是在make.sh中添加
export CXXFLAGS="-std=c++11"
export CFLAGS="-std=c99"
第二點就是采用pytorch0.4.0而不是pytorch0.4.1,然后再make，沒看到報錯，我就假定沒問題了。
接下來琢磨如何訓練自己的數據，官方代碼給的示例是coco2017數據集，這個數據集很大的，幾十G，在網上找了一篇博客https://blog.csdn.net/daniaokuye/article/details/78699138#commentsedit，用wget下載，速度尚可。
下載好后解壓，幾個文件夾如下所示
其中，三個注釋壓縮文件解壓后的默認文件名都是annotations，所以出現了（2）（3），其中train2017中有118287張圖片，19G，val2017有5000張圖片，800多M，test2017有40670張圖片，6.6G。我不太明白的是stuff_annotations里面的注釋文件，查了一下好像是用來做背景語義分割的，做實例分割應該用不到，暫且不管它了，在annotations_trainval2017中存放的是我們想要的語義分割注釋，當然了，里面還有場景描述和人體關鍵點檢測的注釋。如下
然后image_info_test2017里面是測試集的注釋，
左邊的應該是測試圖片的注釋，里面images部分的數目與測試集相等，不過其中只有圖像的信息，並沒有實例的標簽，而右邊的不知道是干嘛的。
現在我准備運行一下提供的測試代碼，不是輸入測試集，而是驗證集，驗證集有注釋文件。
python tools/test_net.py --dataset coco2017 --cfg configs/panet/e2e_panet_R-50-FPN_2x_mask.yaml --load_ckpt {path/to/your/checkpoint
這里要測試的是coco2017驗證集，在主目錄下新建data文件夾，data里面新建coco文件夾，coco里面新建images文件夾和annotations文件夾，images里面放置val2017數據集，annotations里面放置instances_val2017.json文件，這個時候還需要預訓練權重，resnet50_caffe.pth，一開始我下載了pytorch的官方實現，resnet50-19c8e357.pth，發現不行的，鍵值不匹配。報錯如下
Traceback (most recent call last):
File "tools/test_net.py", line 112, in <module>
check_expected_results=True)
File "/home/ustc/文檔/PANet-master/lib/core/test_engine.py", line 128, in run_inference
all_results = result_getter()
File "/home/ustc/文檔/PANet-master/lib/core/test_engine.py", line 108, in result_getter
multi_gpu=multi_gpu_testing
File "/home/ustc/文檔/PANet-master/lib/core/test_engine.py", line 158, in test_net_on_dataset
args, dataset_name, proposal_file, output_dir, gpu_id=gpu_id
File "/home/ustc/文檔/PANet-master/lib/core/test_engine.py", line 232, in test_net
model = initialize_model_from_cfg(args, gpu_id=gpu_id)
File "/home/ustc/文檔/PANet-master/lib/core/test_engine.py", line 322, in initialize_model_from_cfg
model = model_builder.Generalized_RCNN()
File "/home/ustc/文檔/PANet-master/lib/modeling/model_builder.py", line 124, in __init__
self._init_modules()
File "/home/ustc/文檔/PANet-master/lib/modeling/model_builder.py", line 128, in _init_modules
resnet_utils.load_pretrained_imagenet_weights(self)
File "/home/ustc/文檔/PANet-master/lib/utils/resnet_weights_helper.py", line 29, in load_pretrained_imagenet_weights
pretrianed_state_dict = convert_state_dict(torch.load(weights_file))
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/serialization.py", line 303, in load
return _load(f, map_location, pickle_module)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/serialization.py", line 454, in _load
return legacy_load(f)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/site-packages/torch/serialization.py", line 383, in legacy_load
tar.extract('storages', path=tmpdir)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/tarfile.py", line 2034, in extract
tarinfo = self.getmember(member)
File "/home/ustc/anaconda3/envs/py36/lib/python3.6/tarfile.py", line 1745, in getmember
raise KeyError("filename %r not found" % name)
KeyError: "filename 'storages' not found"
於是乎下載resnet50_caffe.pth，搜了一下沒發現什么下載鏈接，因為這個項目利用了detectron.pytorch這個項目的代碼，所以我決定利用detectron.pytorch下載模型權重，項目地址如下https://github.com/roytseng-tw/Detectron.pytorch
安裝了幾個包后，我執行python download_imagenet_weights.py --targets=resnet50_caffe.pth，過了一會，文件下載好了，太開心了，跑起來
python tools/test_net.py --dataset coco2017 --cfg configs/panet/e2e_panet_R-50-FPN_2x_mask.yaml --load_ckpt /home/ustc/文檔/PANet-master/data/pretrained_model/panet_mask_step179999.pth
后面的權重文件是從項目提供的鏈接下載的，一切妥當，開始對5000張圖片進行檢測
然后分別執行bbox評估和分割評估，如下
最后的輸出為評估結果
接下來跑一下coco2017的訓練集，如果沒啥問題，就可以着手跑自己的數據了。
python tools/train_net_step.py --dataset coco2017 --cfg configs/panet/e2e_panet_R-50-FPN_2x_mask.yaml --bs 2
我用的這台機器只有一塊1080ti，所以把把batch_size設為2，論文里是8塊P40，羡慕啊。
訓練跑起來后的樣子如下
可以看到還需要13天訓練完畢，此刻內心是平淡的～
下面就研究如何訓練自己的數據了，我采取替換法，就是文件夾文件名啥的都不換，直接把里面的數據換成自己的，把train2017里面的圖片換成自己的，把instances_train2017.json里面的注釋文件換成自己的，把圖片換成自己的，復制粘貼就行了，不過你的標注文件不一定與coco的格式一致，所以這塊是麻煩的地方。
我采用的標注工具是VIA，使用說明網上有不少博客，如何將VIA格式轉換為COCO格式呢，需要仔細看一下二者的區別，
VIA格式
用json存儲，這一點與COCO相同，json就是一個大字典，VIA采用的鍵值為圖像名稱加一個size(這個size我沒搞清楚，猜測是所有mask的面積和)，圖像名稱下面仍然為一個字典，共有filename,size,regions,file_attributes四個鍵，文件名就是圖像名，size按照我猜測的為掩碼面積和，regions里面就是一個一個mask，file_attribute我沒有用到，置為空。截圖看一下
regions下面存放着許多region,每個region有shape_attributes和region_attributes兩個屬性，一個存放標注的多邊形，一個存放類別，region就是關鍵。
COCO格式
關於COCO的格式，太多太多博客了，就大致看一下吧
總的框架
其中info,licenses都是無關緊要的
對於我們來說，需要關注的是images,annotatios還有categories，分別看一下
對於我們而言，id是可以自行生成的，類別是手動添加的，逐個累加就好，需要計算的是area，bbox，因為VIA格式中沒有這兩個屬性，另外，我的應用場景中，物體都是單一的，所以crowd都設為0，這一點可能需要注意一下
然后就開始具體的轉換了，我把代碼放在了 https://github.com/codingwolfman/VIA2COCO
將VIA轉為COCO后，就可以照常跑起來了。
