Add gridmask and random erasing methods

add fleet training

* Add cutmix op(#88)

* filter illegal data

* update warning message

* optimize python path

* add yolov4

* Make logger less noisy

* Fix indent

* code for cornernet

* add doc

* refine custom op

* refine code from review comments

* refine doc

* update doc

* update code

* add cornernet_squeeze_mixup_cosine config

* update code

* update code

* fcos initialization

* delete the codes not used

* pre-commit the committed code

* delete the unused function

* modify the capacity in loader from 64 to 16

* add mask operations in RandomExpand and RandomCrop

* remove all [] in cropped polys

* is_poly -> _is_poly()

* change the judgment on whether there are polygons

* abstract is_poly to op_helper.py

* add is_mask_expand and is_mask_crop in RandomExpand and RamdomCrop

* add mask operations in RandomExpand and RandomCrop

* remove all [] in cropped polys

* is_poly -> _is_poly()

* change the judgment on whether there are polygons

* abstract is_poly to op_helper.py

* add devices_num on reader

* refine capacity value

* refine bufsize

* add autoaugment, which is validated on faster-rcnn

* enable data unittest

* exit1 in travis failed

* fix all documentation

* fix docs conf.py

* fix some error

* fix readthedocs

* refine cocoapi import

* fix infer when anno_path is None

* split yolov3 loss

* Polish Reader to simplify preprocessing logic.
#  sample_transforms-> make batch -> batch_transforms in Reader.
* Clean some code
* Imporve yolov3_r50vd_dcn_obj365_pretrained_coco 41.4 to 41.8.
* Update all configs.

* Update x2coco.py to fix the cityscape bug when split dataset  

Fix hang bug in parallel_map.py when multiprocessing mode used and some consumer exit abnormally (#106)

* Update GETTING_STARTED_cn.md
* Update x2coco.py

* fix mixup error when dataset has only 1 sample

* refine log

* change fluid.layers.data to fluid.data

* fix export_model doc