* Code for paper "Dynamic R-CNN: Towards High Quality Object Detection via Dynamic Training"

* update configs/dynamic_rcnn/dynamic_rcnn_r50_fpn_1x.py

* reformat code

* simplify code

* update model link

* simplify code

* simplify code logic

* simplify code and add comments

* minor updates of the docstring

Co-authored-by: Jiarui XU <xvjiarui0826@gmail.com>
Co-authored-by: Kai Chen <chenkaidev@gmail.com>

* Add docstrings of core modules and methods

* Update docs and fix comments

* Complete docstrings

* Resolve comments

* reformat docstrings

* resolve comments

Rename faster_rcnn_regnetx-3GF_fpn_mstrxin_3x_coco.py to faster_rcnn_regnetx-3GF_fpn_mstrain_3x_coco.py (#3126)

Fix potential typo

* add voc benchmark

* add faster rcnn voc res

* Refactor GRoI to support concat mode

* add base roi extractor

* fixed num_inputs property

* update test

* add docstring

* update doc string and __init__ order

* update doc

* fixed test

* add default args

* clean legacy deprecated usage of register_module

* Update doc

* Use len(data['img_metas']) to indicate num_samples

* clean code

* switch to EpochBasedRunner

* add docstrings

* update the minimum version of mmcv to 0.6.0

* fix unit tests

* fix mmcv version in travis

* add mmfashion dataset

* Update __init__.py

* Update mmfashion.py

* rename mmfashion to DeepFashion

* refactor deepfashion config file

* remove whitespace

* refactor deep fashion config file

* refactor deepfashion config file

* add a new line

* add a blank line

* upload log

* omit default setting in config

* rename deepfashion config

Co-authored-by: Jiaqi Wang <wj017@ie.cuhk.edu.hk>

* 123

* implement FCOS with tricks (FCOS_plus)

Train FCOS with four tricks (including center_sampling) arrording to official repo (https://github.com/tianzhi0549/FCOS).

without DCN:  mAP=38.6 (0.1 lower than official repo)
with DCN     :  mAP=42.5 (0.2 higher than official repo)

* fix

* fix

* reformatted

* fix configs & reformat

* add docstring

* Update fcos_center-normbbox-centeronreg-giou_r50_caffe_fpn_gn-head_dcn_4x4_1x_coco.py

* add noqa: E501 since the url is too long

* fix flake

* fix flake

* fix yapf

* fix

* rewrite forward & reformat

* reformat & add docstring

* reformat

* reformat

* update docstring

* TODO: add unit test

Should be finished tomorrow

* add unit test

* fix

* fix docstring

* reformat

* add docstring

* reformat

* Update fcos_head.py

* Update fcos_head.py

* Update fcos_head.py

* Update fcos_head.py

* Update fcos_head.py

* Update test_heads.py

* Update fcos_head.py

* Update README.md

* Update README.md

* 123

* implement FCOS with tricks (FCOS_plus)

Train FCOS with four tricks (including center_sampling) arrording to official repo (https://github.com/tianzhi0549/FCOS).

without DCN:  mAP=38.6 (0.1 lower than official repo)
with DCN     :  mAP=42.5 (0.2 higher than official repo)

* fix

* fix

* reformatted

* fix configs & reformat

* add docstring

* Update fcos_center-normbbox-centeronreg-giou_r50_caffe_fpn_gn-head_dcn_4x4_1x_coco.py

* add noqa: E501 since the url is too long

* fix flake

* fix flake

* fix yapf

* fix

* rewrite forward & reformat

* reformat & add docstring

* reformat

* reformat

* update docstring

* TODO: add unit test

Should be finished tomorrow

* add unit test

* fix

* fix docstring

* reformat

* add docstring

* reformat

* Update fcos_head.py

* Update fcos_head.py

* Update fcos_head.py

* Update fcos_head.py

* Update fcos_head.py

* Update test_heads.py

* Update fcos_head.py

* Update README.md

* add ae loss and gaussian focal loss

* update losses/__init__.py

* formatting

* Update __init__.py

* update losses

* add weighted_loss in gaussian_focal_loss

* rename AELoss, move comment to docstring

* fix formatting

`target_means` and `target_stds` should be in bbox_coder according to
https://github.com/open-mmlab/mmdetection/blob/2b6f6616f804beaca3dbf071fa398c586243db13/configs/_base_/models/mask_rcnn_r50_fpn.py#L48

* use BaseDenseHead

* use RPNTestMixin in rpn heads

* use self(x) instead of self.__call__(x)

* use kwargs

* add Hourglass and unittest

* code formatting

* fix bug

* rename and add docstring

* add unittest for ResLayer(reverse=True)

* update Hourglass

* update Hourglass, update unittest

* update docstrings, rename variables

* update hourglass and unittest

* update hourglass.py and res_layer.py

* Fixed nan loss by filtering out-of-frame gt_bboxes

* Discarded lambda expression of is_out_of_frame

* Cleaned trailing whitespace

* Reformatted code

* Checked the intersection between boxes and image

* save for debug

* add foreground dominate

* debug modify

* debug for mmdet

* small change

* small change

* bug fix

* bug fix

* update project link

* Update projects

* Add reimplementation template

* reformat

* reformat

* Fix typos

* resolve comments

* Modify mdconv doc

* add todo

* undo

* add fl num class check

* format

* remove space

* add space

Edit typographical errors.

* update project link

* Update projects

* add with_cp for basicblock of resnet, fixes #2887

* reformat code

* fix bugs

sorry but github edit is so hard to use......

* fix unittests

Co-authored-by: Kai Chen <chenkaidev@gmail.com>

* add center pooling

* fix yapf formatting

* add unittest

* fix flake8 formatting

* fix isort formatting

* fix isort formatting

* refactor and formatting

* remove top pool from unittext

* remove bottom pool from unittest

* fix bug in torch1.5

* add pytest for assertion

* center_pool -> corner_pool, add docstring

* fix unittest

* add corner_pool.py

* fix some weird logics

* move unittests of ops to a new folder

Co-authored-by: Kai Chen <chenkaidev@gmail.com>