* 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

* rename BitMap to Bitmap, add input check for polygon

* fix test mask

* fix test config

* complete test

* add mask contest test for bitmap resize

* update with np.diag

* perfect test polygon resize

* perfect test polygon crop

* add CPU only mode which can be activated during install

* fixed flake8 errors for too long lines, still have to deal with "import not at top of file"

* reversing changes in MinIoURandomCrop, that not relevant to the CPU_ONLY pull request

* moving the CPU_ONLY checks into deeper parts of the code

* completing previous commit

* using isort for imports sorting

* yapf fix

* followed @xvjiarui suggestions for the pull request

* use mmdet.CPU_ONLY and replace the "--cpu" flag in setup.py with automatic check of cuda

* make setup code cleaner

* back to original implementation of MinIoURandomCrop

* build all extensions with CUDA, if available

* fixed DC

* update doc

* fixed masked_conv2d_ext

* set warning once, update comment

Co-authored-by: Yossi Biton <yossi.biton@alibaba-inc.com>
Co-authored-by: Yossi Biton <yossibit10@gmail.com>

* refactor mask structure

* update and add comment

* update test_config, add crop_and_resize

* update to comment

* add unittest

* Update .travis.yml

* add iter test

* check mask type in albu

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

* add merge conv bn

* fix wrong push

* support save merged model

* rename merge to fuse, add annotations

* fix typos

* Feature (smooth-l1): support to use L1 loss

* Refactor (L1-loss): Create a independent class for L1-loss

* Refactor (l1-loss): add l1 loss in __init__

* Refactor (L1-loss): rm kwargs

* 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

* Modify the code casting digit string as int to convert to float and then to int

* Add Comment: Coordinates may be float type

* Fix: Indentation Error

* refactor gcnet/emperical into plugin

* add missing plugin

* add docs, make plugins stage more flexible

* refactor plugins into list[dict]

* add duplicate test, update docs

* update docs

* update docs

* update comment

* update testbackbone doc

* create roi_head

* create roi_head
mv shared_head into roi_head

* fix conflict

* refactor(cascade rcnn): create cascade head in roi head, simplify cascade rcnn

* type: fix
scope: cascade_head
subject: mv stage_loss_weight from train_cfg to roi_head creation

* type: fix
scope: roi_heads
subject: handle the case when train_cfg is None

* type: fix
scope: roi_heads mask inference
subject: fix test_cfg.rcnn bug to test_cfg

* fix mask eval bug

* refactor(double_head): refactor double head roi head

* fix(double_head): clean code and fix __init__.py

* refactor(mask scoring): create mask scoring roi head

* refactor(htc_head): add htc_head

* fix(htc_head): add htc_head in __init__.py

* fix (htc_head): fix htc_head bugs

* fix (ms_roi_head): rm test_cfg.rcnn to test_cfg

* feature (grid rcnn): grid roi head

* fix (grid_head): fix grid head bug

* reformat and change all cfgs

* reformat (roi_head): reformat __init__.py for isort=4.3.21

* fix roi_head test bug

* fix (carafe): change carafe cfgs to use roi_head

* fix (roi_head): fix missing roi_head refactor

* reformat to pass CI

* test all cfgs

* match keys of configs with previous commit 77d073a

* add and pass unittest for all roi heads

* Refactor (roi_head): extract abstract base class for roi head

* Refactor (roi_head): refactor init functions

* Refactor (roi_head): weight init

* Refactor (roi_head): add _bbox_forwrd & _mask_forward as basic functions

* Fix (grid_roi_head): fix bug in bbox_forward_train

* Refactor (roi_head): change to use img.device in forward_dummy)

* Refactor (roi_heads): simplify init functions and _mask_forward

* Fix (cascade_roi_head): fix init bug of cascade roi_head

* Refactor (roi_head): use dict as outputs of _bbox_forward and _mask_forward

* Refactor (test_config): scan valid configs rather than list them all

* Fix (flip): not setting default values of flip during loading

* Fix (loading): correct the format of default img_norm_cfg

* add condition on RoIPool backward

* format

* delete space

* reformat

* add newline end of file

* refactor resnet, add deep_stem and avg_down opt

* move args into kwargs

* add default

* mv resnetv1d to resnet

* fixed res_layer

* add base_channels

* add doc

* add tests

* update test

* fixed backbone block conv_cfg/norm_cfg

* add test

* add tests

* add unit tests for image loading

* add test data

* add a test case

* add norm_eval

* upgrade mmcv

* delete work_dir in all configs

* add comment on work_dir priority

* add new config sample and check scripts

* context managere to prevent import

* refactor backbone, add caffe

* add fast rcnn

* all basci config done

* rm coco_proposal

* add scratch, wider_face, voc

* add reppoints

* delete device id, auto resume

* add libra_rcnn, ms_rcnn, nas_fpn

* add hrnet, fcos

* add free_anchor, gn, grid_rcnn, guided_anchoring, gcnet, ghm, gn+ws

* finished all, but some need refactor and unified logic

* fixed albu

* update ReseNet101

* reduce ResNet101, add norm_cfg, 4conv1fc

* refactor caffe backbone

* merge act_cfg, dcn from master

* rename 2fc head, add 4conv1fc head

* delete 4con1fc model

* rename components -> _base_

* flatten _base_

* refactor _base_ structure

* refactor config structure

* reduce _base/models

* rename all config

* add _coco inside configs

* fixed _coco in files

* fixed albu in test

* migrate to new sys

* fixed test

* fixed test

* rm cascade, fast_rcnn

* rm cascade, fast_rcnn from _base_

* fixed fcos_center

* rename from rule

* collopse third party model

* fixed test

* fixed some minor bug

* rename gn-neck_head -> gn-neck+head

* 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 bug: index out of range when num_extra < len(indice)

* Unified img_metas api

* minor fix

* fix typo

* update images for async test

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

* Add class to wrap result fields into lists for evaluation

On branch wrap_into_lists
Changes to be committed:
modified:   mmdet/datasets/pipelines/formating.py

* Add doctsring to WrapFieldsToLists

On branch wrap_into_lists
Changes to be committed:
modified:   mmdet/datasets/pipelines/formating.py

* update the example in docstring of WrapFieldsToLists

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

* fix bug of mse_loss

* reformat

* use single quote

Co-authored-by: penghaozhou <penghaozhou@tencent.com>
Co-authored-by: wwzhang <wayne.zw@outlook.com>

On branch load_multichannel
Changes to be committed:
modified:   mmdet/datasets/pipelines/loading.py

* fix dcn when input size smaller than kernel size

* update

In example 4:

Test Mask R-CNN on COCO test-dev with 8 GPUs, and generate the json file to be submit to the official evaluation server.
./tools/dist_test.sh configs/mask_rcnn_r50_fpn_1x.py
checkpoints/mask_rcnn_r50_fpn_1x_20181010-069fa190.pth
8 --format-only --options "jsonfile_prefix=./mask_rcnn_test-dev_results"

where --format-only should be --format_only

* Set initial values for default meta keys when loading images

On branch set_default_metakeys
Changes to be committed:
modified:   mmdet/datasets/pipelines/loading.py

* Change local variable name

On branch set_default_metakeys
Changes to be committed:
modified:   mmdet/datasets/pipelines/loading.py

On branch list_to_kvaction
Changes to be committed:
modified:   tools/test.py

* fix (transforms): fix bugs in transforms when the gt is empty

* Fix (SSD): fix bug of data pipeline used in SSD, test more pipelines

* Fix (test_config): move print for more clear debug info

* clean comments

* add onnx convert script

* upload pytorch2onnx.py

* restore config

* set use_torchvion on-the-fly, update doc

* update doc

* support pass input

* fixed passes