* Update pytorch2onnx.py which using new logic to convert pytorch to ONNX

* use standard API to check whether in ONNX convert process

* only compare the score value while verifying results between ONNX and pytorch

* move import onnx before import torch, or something weird will happen

* use real images for input

* modifying the way of calling nms

* modify docstring for bbox2result, and remove unnecessary part for onnx exporting

* modify the 'Convert to ONNX' part in docs

* replace or to | in docstring

* update according to the latest mmcv

* add normalize part

* raise error while using low version mmcv

* minor update

* minor update

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

* add pat change

* fix grid roi head

* fix comments

* clean

* revert change

* update FSAF model zoo

* Add foreground_dominate argument to center region assigner

* Refine docs and fix format issues

* reformat

* fix mock

* Refine format

* Reorganize tutorials

* clean op

* clean

* update build.yml

* fix setup

* change ci mmcv version

* change ci mmcv version

* add install pre-build mmcv instructions

* update mmcv url, move some unnit tests to mmcv

* update ci mmcv

* test

* test

* test

* fix unit test

* fix unit test

* fix onnx script

* update mmcv url

* fix typo

* fix some doc

* Bump version

* resolve comments

* Fix version bug

Co-authored-by: ZwwWayne <wayne.zw@outlook.com>

* processing ignore boxes in atss assigner

* formatting

* flake8 correct

* correct flake8 v2

* frommatting via pycharm

* OMG this is the last commit

* fix bug

* format

Co-authored-by: d.rybalchenko <d.rybalchenko@visionlabs.ru>
Co-authored-by: Cao Yuhang <yhcao6@gmail.com>

* update docstring formats

* fix :func:

* minor update

* add gfocal losses

* add gfl head, its related code, configs

* fix 2x config bug

* add most model and log in gfl README

* update gfl model and log

* Update transforms.py

modify variables  L, T, R, B   to  left, top, right, bottom

* Update gfocal_loss.py

modify variables into meaningful names:
* pt -> scale_factor
* a -> pos   b -> pos_label
* disl -> dis_left  disr -> dis_right
* wl -> weight_left  wr -> weight_right

* Update gfocal_loss.py

update formats

* remove duplicated get_bboxes/IoU calculation, make QFL consistent

* add all docstring in gfl_head.py and gfocal_loss.py

* Add docstrings of core modules and methods

* Update docs and fix comments

* Complete docstrings

* Resolve comments

* reformat docstrings

* resolve comments

* fixed vertical flip TTA

* fixed test

* add doc and one more check

* fixed docs

* save

* add retinanet and fix some inconsistent

* fix registry

* fix typos

* fix unmap bug

* fix some bugs

* clean

* minor fix

* fix k b

* add more cfgs

* fix detach

* add some benchmarks

* add some comments

* add more comments

* change test rpn proposal num

* add more models

* modify accroding comments

* more fix

* add pape link, fix some grammar error

* modify according to comments

* add unit test

* add unit tests for head

* fix some comments, add pape link

* add pisa roi head unittest

* fix README.md

* fix year

* add api doc

* update mock modules and fix requirements

* fix autodoc modules

* remove tmp file

* add release version in docs/conf.py

* fix docstring styles

* migrate to mmcv.cnn.bricks

* remove conv_module, scale, upsample

* update mmdet version and mmcv requirement

* fixed fsaf warning

* starting to merge fsaf

* changed configs

* changed bbox_coder

* get_targets fun in AnchorHead class is modified to enable additional returns from _get_target_single

* effective area assigner and tblr bbox_coder are implemented

* added fsaf_head and config

* added more fsaf configs

* fixed the nan bbox_loss by clamping bbox_pred with min=1e-4

* fixed the nan bbox_loss by clamping bbox_pred with min=1e-4

* changed config

* formatter

* changed docstring

* Added ignored pixel-class pair for FSAF

* More elegant assignment for overlapping objects

* more documents in effective area assigner

* fixed a bug in object overlay

* Added no gt_bboxes case and pytest for assigner

* Added no gt_bboxes case and pytest for assigner

* changed default config

* fixed small bugs in FSAF head doc example

* resolved pr review comments.
1. Formats of comments and docstrings are modified
2. Obsolete arguments are removed from FSAFHead
3. AssignResults class is revised to be more compatible with old version

* Added more comments to FSAFHead

* resolved pr review comments

* resolved pr review comments
1. More flexible CenterRegionAssigner bbox priority
2. More flexible tblr_coder by adding normalize_by_wh flag
3. README modifications with model link
4. More docstrings and comments
5. Reformatted
6. Change of default normalizer in TBLRBBoxCoder from 1.0 to 4.0

* Corrected FSAF model path in readme

* Fixed the bug of bbox weights in fsaf head

Co-authored-by: wangxinjiang <wangxinjiang@sensetime.com>

* add model upgrade tool

* Add configs and scripts for converting v1.x models to v2.0

* add doc

* Add more models

* Pass CI

* Clean scripts and check model by config

* Resolve comments

* Resolve small comments

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

* rm bboxes[:,:4]

* Move the bbox size selection and check into iou2d_calculator

* Add bbox shape check in bbox_overlaps

* Validate the bbox sizes

* added background label to assign function

* fixed the background label as -1 in all assigners

Co-authored-by: wangxinjiang <wangxinjiang@sensetime.com>

* Change .format to f-string

* Resolve comments

* Fix missing change

* use mmcv Registry instead of mmdet

* merged registry and builder

* merge dataset builder registry

* remove circulate import in bbox

* rename registry to builder

* fixed init/refine

* fixed bbox coder assert

* add reg_decoded_bbox in two_stage

* finished single stage

* fixed comment

* fixed docs

* add coder registry

* add coder for single stage

* rename to bbox_coder

* add doc string

* add doc string

* change default

* merge target into head

* remove squeeze in image_to_level, add doc string

* reformat doc string

* update docs, rename to get_target

* reanme unify all target functions to get_targets

* set test_cfg as default cfg for get_bboxes

* fixed test

* rename get_targets in atss, fcos ,fovea, reppoint

* rename get_det_bboxes to get_bboxes

* Refactor (assigner): add iou_calculator

* Rename (iou_calculator): rename registry IOUCALCULATOR to  IOU_CALCULATOR

* resolve comments

* Rename IOU_CALCULATOR to IOU_CALCULATORS

* add sampler and assigner registry

* rename with bbox prefix

* restore __init__

* roll back atss_head

* change import level

* import from sampler/assigner

* Refactor (all): all category -1 in anchor/bbox head and anchor/bbox target assign

* Fix (datasets): remove label + 1 in datasets

* Fix (bbox_head): fix bug of fc_cls that forget + 1

* Fix (atss_head & free_anchor): fix cat -1 bugs

* Fix (mask_head): remove label + 1 in mask heads

* fix atss

* Fix (rpn): fix cross_entropy_loss bug of RPN

* Fix (anchor_head): fix typo

* Refactor (anchor_head): use background_label rather than num_classes to indicate background class

* Refactor (docstring): add and reformat docstrings

* fix mask iou head

* Fix (mask_head): fix cat -1 bug

* Fix (mask_head): fix bug in mask inference

* Add (tests): add tests for mask rcnn and mask heads

* Refactor (unittest): refactor test_forward

* Refactor (new_empty): use new_full rather than new_empty

* Refactor (background_label): check background_label

* Add TODO

* Refactor (unittest): allow BP in unittest

Co-authored-by: Cao Yuhang <yhcao6@gmail.com>

* Refactor (all): change coordinate system

* Fix (mask_head): fix cat -1 bug in mask_paste

* Fix (unittest)
: modify unittest and pass CI

* reformat to pass CI

* Fix round coordinates bugs

* clean file

* Fix (test): use cpu version of aligned roi_align in tests

* Refactor (mask): clean np.stack

* Refactor (head): reformat code and fix missing -1

* Reformat: reformat and add doc strings

* Refactor (mask_head): more clea docstring

* Refactor (iou_assigner): set low quality match as an option

* Fix (max_iou_assigner): fix typo and add clear detail

* Fix (cascade_rcnn): set match_low_quality in cascade rcnn

* fix (scale_factor): use new_tensor(scale_factor) in case it is numpy array

* reformat (models): reformat with flake8, yapf, and isort to pass CI

* fix (bbox_mapping): fix scale_factor bug in bbox_mapping

* fix img_meta bug

Co-authored-by: beansi <zhangwenwei@sensetime.com>

* Fix IOU assigners when ignore_of_thr > 0 and no pred boxes

* remove extra brakets

* fix test linter

* speed up sampler

* remove unused numpy

* add complete doc