reuse RetinaNet pipeline for now， force merge to prevent CI block

* update link on README

* use relative path

* Add PrepareData-->SelectModel-->Train-->Eval-->Test-->Infer-->Slim-->Deploy document DetectionPipeline.md, test=document_fix

* modify cpp output path

* fix compile error

* modify docs for windows

* share train_batch_size from reader to model

* update floor_divide

* add makedirs for dumping config

* check md5 after weights download

* move train_batch_size sync from load_config to merge_config

* Add install COCO-API method on windows system, test=document_fix

* Add tips, test=document_fix

* Change demo image, test=document_fix

* warning when YOLOv3Loss.batch_size setting

* add batch_size in YOLOv3Loss for backward compatable.

* sync TrainReader.batch to global train_batch_size

* cherry-pick #1375 yolov3_mobilenet_v1.md

* add link, test=document_fix

* cherry-pick #1406,#1407,#1420,#1422

* add usage of tools/anchor_cluster.py in doc (#1440)

* modify doc for anchor cluster

* fix voc mAP %.

* add tools/anchor_cluster.py

test=document_fix

* Update CN README
* Change demo and update doc

Co-authored-by: Nqingqing01 <dangqingqing@baidu.com>

Co-authored-by: NKaipeng Deng <dengkaipeng@baidu.com>

* add roadsign_voc dataset

* fix some typo