Saving data dict to COCO format

Thanks for the excellent work! I’ve encountered some issues around saving my own annotations into COCO format.

How To Reproduce the Issue

from detectron2.data.datasets.coco import convert_to_coco_json
from detectron2.data import MetadataCatalog, DatasetCatalog
def data_dict():
    return [{
        'file_name': 'image.png',
        'image_id': 'image',
        'height': 100,
        'width': 100,
        'annotations': [{
            'bbox': [70, 30, 100, 70],
            ## area should be 30*40=1200
            'bbox_mode': BoxMode.XYXY_ABS,
            'category_id': 0,
            'iscrowd': 0,
        }]
    }]

def to_json():
    DatasetCatalog.register('test', data_dict)
    MetadataCatalog.get('test').set(thing_classes=["first"])
    convert_to_coco_json('test', output_folder='./output', allow_cached=False)

if __name__ == '__main__':
    to_json()

After running the script above, in ./output/test_coco_format.json I get: (formatted)

{
    "info": {
        "date_created": "2019-11-19 23:23:08.401029",
        "description": "Automatically generated COCO json file for Detectron2."
    },
    "images": [
        {
            "id": "image",
            "width": 100,
            "height": 100,
            "file_name": "image.png"
        }
    ],
    "annotations": [
        {
            "id": 1,
            "image_id": "image",
            "bbox": [
                70.0,
                30.0,
                30.0,
                40.0
            ],
            "area": -400.0,
            "category_id": 0,
            "iscrowd": 0
        }
    ],
    "categories": [
        {
            "id": 0,
            "name": "first"
        }
    ],
    "licenses": null
}

Expected behavior

The area in annotation should be positive (1200) I think the problem is here, but I’m not sure how to correct it. The .area() function isn’t considering its BoxMode.
BTW, Is saving my own annotation into COCO format and registering it by register_coco_instances the best practice when using COCO_Evaluator?

Environment

------------------------  --------------------------------------------------
sys.platform              linux
Python                    3.7.3 (default, Mar 27 2019, 22:11:17) [GCC 7.3.0]
Numpy                     1.17.4
Detectron2 Compiler       GCC 7.4
Detectron2 CUDA Compiler  10.0
DETECTRON2_ENV_MODULE     <not set>
PyTorch                   1.3.1
PyTorch Debug Build       False
torchvision               0.4.2
CUDA available            True
GPU 0,1,2,3               GeForce RTX 2080 Ti
CUDA_HOME                 /usr/local/cuda-10.0
NVCC                      Cuda compilation tools, release 10.0, V10.0.130
Pillow                    6.2.1
cv2                       4.1.1
------------------------  --------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - Intel(R) Math Kernel Library Version 2019.0.4 Product Build 20190411 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v0.20.5 (Git Hash 0125f28c61c1f822fd48570b4c1066f96fcb9b2e)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CUDA Runtime 10.1
  - 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_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_37,code=compute_37
  - CuDNN 7.6.3
  - Magma 2.5.1
  - Build settings: BLAS=MKL, BUILD_NAMEDTENSOR=OFF, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -fopenmp -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -O2 -fPIC -Wno-narrowing -Wall -Wextra -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-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -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 -Wno-stringop-overflow, DISABLE_NUMA=1, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=True, 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, USE_STATIC_DISPATCH=OFF, 

[botcs]

Hi, nice catch!

@ppwwyyxx BoxMode should be considered in https://github.com/facebookresearch/detectron2/blob/master/detectron2/structures/boxes.py#L118 not in convert_to_coco_json

@donnydonny123 if you do not use external tools that require your data to be in COCO format it is indeed quite pointless.

[ppwwyyxx]

Yes, the Boxes class is defined to be in XYXY_ABS mode only. So a conversion before constructing Boxes is needed.