rotated faster rcnn offset

Full runnable code or full changes you made:

Hello, I am running rotated faster rcnn with detectron2 in radar dataset called radiate. radiate: https://github.com/marcelsheeny/radiate_sdk And I used this code.

(https://github.com/marcelsheeny/radiate_sdk/blob/master/vehicle_detection/train.py)
Also yaml file is below. (faster_rrcnn_R_50_FPN_3x.yaml) I think they called rotated rcnn to rrcnn.

BASE: "Base-RCNN-FPN.yaml"
MODEL:
PIXEL_MEAN: [103.530, 116.280, 123.675] # Pixel mean of imagenet dataset R/G/B?
WEIGHTS: "detectron2://ImageNetPretrained/MSRA/R-50.pkl"
MASK_ON: False
RESNETS:
DEPTH: 50
PROPOSAL_GENERATOR:
NAME: "RRPN"
RPN:
HEAD_NAME: "StandardRPNHead"
BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0, 1.0)
ANCHOR_GENERATOR:
NAME: "RotatedAnchorGenerator"
ANGLES: [[0, 30, 60, 90]]
ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
ROI_HEADS:
NAME: "RROIHeads"
NUM_CLASSES: 1 # number of foreground classes
ROI_BOX_HEAD:
POOLER_TYPE: "ROIAlignRotated"
BBOX_REG_WEIGHTS: (10.0, 10.0, 5.0, 5.0, 1.0)
SOLVER:
STEPS: (210000, 250000)

What exact command you run: trained ok
Full logs or other relevant observations:

When inference is made based on the learning result (AP50: 23), a bounding box offset like the image below occurs.

https://user-images.githubusercontent.com/79086973/148164317-4b33b14c-54f1-4eac-8518-7bb5cfde2251.mp4

It was confirmed that the label of the dataset itself was well done. When running rotated rcnn on detectron2 code, is the offset reflected?

Please advise on how to solve the problem.

not log

Environment:

sys.platform            linux
Python                  3.7.11 (default, Jul 27 2021, 14:32:16) [GCC 7.5.0]
numpy                   1.21.4
detectron2              0.6 @/data1/Dongin/radiate_sdk/detectron2/detectron2
Compiler                GCC 7.5
CUDA compiler           CUDA 11.2
detectron2 arch flags   8.6
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.9.0+cu111 @/home/oem/anaconda3/envs/donginws/lib/python3.7/site-packages/torch
PyTorch debug build     False
GPU available           Yes
GPU 0,1,2,3             GeForce RTX 3090 (arch=8.6)
Driver version          460.91.03
CUDA_HOME               /usr/local/cuda
Pillow                  8.4.0
torchvision             0.10.0+cu111 @/home/oem/anaconda3/envs/donginws/lib/python3.7/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5, 8.0, 8.6
fvcore                  0.1.5.post20211023
iopath                  0.1.9
cv2                     4.5.4
----------------------  ---------------------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v2.1.2 (Git Hash 98be7e8afa711dc9b66c8ff3504129cb82013cdb)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 11.1
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86
  - CuDNN 8.0.5
  - Magma 2.5.2
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.1, CUDNN_VERSION=8.0.5, CXX_COMPILER=/opt/rh/devtoolset-7/root/usr/bin/c++, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_KINETO -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.9.0, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

Testing NCCL connectivity ... this should not hang.
NCCL succeeded.

facebookresearch / detectron2

rotated faster rcnn offset #3853

Environment: