Summary:
When 'model_final.pkl' exists, the following error occurs:
```shell
INFO train_net.py: 184: model_final.pkl exists; no need to train!
Traceback (most recent call last):
  File "tools/train_net.py", line 280, in <module>
    main()
  File "tools/train_net.py", line 119, in main
    checkpoints = train_model()
  File "tools/train_net.py", line 128, in train_model
    model, weights_file, start_iter, checkpoints, output_dir = create_model()
ValueError: need more than 4 values to unpack
```
It misses a return value.
Closes https://github.com/facebookresearch/Detectron/pull/309

Reviewed By: ir413

Differential Revision: D7366733

Pulled By: rbgirshick

fbshipit-source-id: c37b3b2e5b2aff38d46ca521a9d79613049255cb

Reviewed By: ir413

Differential Revision: D7364656

fbshipit-source-id: bfd31bc7c95b9606037c2f5546c9edd0e0318272

Reviewed By: ashwinb

Differential Revision: D7148434

fbshipit-source-id: 4f5e6386d484b0012ed14dbe1445ee8010d2cc26

Reviewed By: ir413

Differential Revision: D7148432

fbshipit-source-id: e23f767fe0241496b0ee0fb73c247747250eb34e

Reviewed By: ir413

Differential Revision: D7148430

fbshipit-source-id: 71f25716d157daf4082a35dd2138debd9a43c7f1

Reviewed By: ir413

Differential Revision: D7148427

fbshipit-source-id: 25b755c75ddc59cff7a1dc5a5cddb139c44f7cbf

Reviewed By: ir413

Differential Revision: D7148425

fbshipit-source-id: 78c11427b0559248116ca1065b8252355f0034a3

Reviewed By: ashwinb

Differential Revision: D7148419

fbshipit-source-id: c8990b7db3c48ea76b665cbda1b9c0985096920c

Summary:
Fix typos when creating symlinks for PASCAL VOC.
Closes https://github.com/facebookresearch/Detectron/pull/275

Reviewed By: rbgirshick

Differential Revision: D7274459

Pulled By: ir413

fbshipit-source-id: 8a60516ac199371f67117b29f1ebe3199dabe550

Reviewed By: rbgirshick

Differential Revision: D7271967

fbshipit-source-id: f42493b08e097bd702b36482b4587a42e5bf3262

Reviewed By: rbgirshick

Differential Revision: D7086631

fbshipit-source-id: 8d1a756bdd187eb33cde0dac0dc73c5293ca17eb

Reviewed By: rbgirshick

Differential Revision: D7084409

fbshipit-source-id: b1fc559847710e7e4dc35fa8b317c0ba259a4dde

Reviewed By: rbgirshick

Differential Revision: D7082759

fbshipit-source-id: b3257e418944dcfb3c15ff71404c39451d4e225d

Reviewed By: rbgirshick

Differential Revision: D7069768

fbshipit-source-id: f254f0073ddc57c43977a8010ca67e89ea1e0c3f

Summary:
Fixed bug preventing RetinaNet training with num_classes != 81 from converging.

Resolves #93
Closes https://github.com/facebookresearch/Detectron/pull/176

Reviewed By: ir413

Differential Revision: D7056134

Pulled By: rbgirshick

fbshipit-source-id: d3c62ce98a23a09191dd2720bc7189ee5490635d

Summary:
This PR fix the issue vis_one_image_opencv raises ValueError.

>Since OpenCV 3.2, findContours() no longer modifies the source image but returns a modified image as the first of three return parameters.
https://docs.opencv.org/3.3.1/d4/d73/tutorial_py_contours_begin.html

and this line also assumes opencv>=3.2
https://github.com/facebookresearch/Detectron/blob/master/lib/utils/vis.py#L326

so, It may be better to set OpenCV version specification to 3.2 or higher in Installation guide.
Closes https://github.com/facebookresearch/Detectron/pull/9

Reviewed By: rbgirshick

Differential Revision: D7009465

Pulled By: ir413

fbshipit-source-id: 3b2cea6c94da1de93927a6d11724f104c063dca1

Summary: Add link to model zoo training and inference logs

Reviewed By: ir413

Differential Revision: D6882050

fbshipit-source-id: 8dabe6ec3b56b47d9f2f00b9d01630eb9911c080

Reviewed By: rbgirshick

Differential Revision: D6875655

fbshipit-source-id: 9c19510ff5c402bef2d947d95d061a5d51e57e9a

Reviewed By: rbgirshick

Differential Revision: D6871605

fbshipit-source-id: 284a9267645cb039c2b20b8ceab68309664b0133

Reviewed By: rbgirshick

Differential Revision: D6824147

fbshipit-source-id: fe06c4b4ee200003d5d9e4ca62e4afc9741cd929

Summary:
Under certain settings, it is desirable to use multi-threaded
evaluation even though it is inefficient due to Python's GIL. This set of
changes allows for it.

Reviewed By: rbgirshick

Differential Revision: D6827125

fbshipit-source-id: 845647bd09198c35bb6d06c1a453a817d0dcf784

Reviewed By: rbgirshick

Differential Revision: D6848494

fbshipit-source-id: 1bc4605443060891872ec1cc7c78216980440003

Summary:
Generalizations to enable building forward passes and param update
operations separately.

Reviewed By: rbgirshick

Differential Revision: D6827027

fbshipit-source-id: 9d2ace349b4ebdf8993baa2f2c1529e2c5ce751d

Reviewed By: rbgirshick

Differential Revision: D6845192

fbshipit-source-id: cf172ab7c9fa7137d35839b721b517576994eb35

Summary:
Under a distributed setting (multiple nodes), the loss scale may need
to account for the number of nodes participating in the algorithm. This method
provides a reasonable way for providing this config.

Reviewed By: rbgirshick

Differential Revision: D6826976

fbshipit-source-id: d224200710b0f3c3245851874c9f4f3ab871710f

Summary:
The `AffineChannel` op wrapper did not use the officially sanctioned
version of parameter creation, namely using `ModelHelper.create_param()`
Presumably this was because of the `share_with` option where you could share
parameters.

While doing this though, the parameters were never initialized! This probably
doesn't matter for the algorithm (those params get overwritten pretty quickly)
but it's a bug nevertheless. I decided to kill the `share_with` option since
this isn't used anyway and resort to the standard way of creating weights and
biases.

Reviewed By: rbgirshick

Differential Revision: D6826914

fbshipit-source-id: 65628c59b085b9ab160006b003dd40dbefa2f7c1

Summary:
This diff starts a series of diffs to refactor the detectron codebase
so it can be used in an elastic data parallel context (see
https://our.intern.facebook.com/intern/dex/caffe2/elastic-data-parallel-model-for-distributed-training/)
without any forks.

Specifically, this diff does the following:

 - Splits out the `TrainingStats` class so it can be reused / composed
 - Slightly refactor of `initialize_weights_from_file()`
 - Reduces the verbosity of some logs

Hopefully, nothing controversial here :)

Reviewed By: rbgirshick

Differential Revision: D6826820

fbshipit-source-id: fc15209a0ff50e5d09281e36173198c77aa77a12

Summary: Revise print_net to show the full forward model when there are multiple losses (which is always the case).

Reviewed By: rbgirshick

Differential Revision: D6829753

fbshipit-source-id: 53c1d63d95a78f83a0275b90cad1517315786e7d

Reviewed By: KaimingHe

Differential Revision: D6842986

fbshipit-source-id: c8370e2f831239b5489dc723738c6bdf5708a40d

Summary:
Added a script to convert pkl model to pb.
- Supported Faster R-CNN detection model.
- Supported converting to CPU and GPU model.
- Fused AffineChannel to Conv.
- Verified model after conversion.

Reviewed By: rbgirshick

Differential Revision: D6783633

fbshipit-source-id: 19706d3074a4a784a2161a695d7c534e014ebb3f

Reviewed By: rbgirshick

Differential Revision: D6829499

fbshipit-source-id: 66fe40ef1b2c8c560ce769faa27d107c3c7be841

Reviewed By: rbgirshick

Differential Revision: D6829350

fbshipit-source-id: a5fbaa8900fab28602d7a993d63206ca67b24975

Reviewed By: rbgirshick

Differential Revision: D6829206

fbshipit-source-id: 06e8917366a68e4c2e44637d784b019ba2d6fddf

Reviewed By: rbgirshick

Differential Revision: D6829095

fbshipit-source-id: 1d36fd7adf1596350017c437a4365d360a21dccd

Reviewed By: rbgirshick

Differential Revision: D6828915

fbshipit-source-id: dfc8bd970c0555d56bfd7ad313550b0a12c154f2

Reviewed By: rbgirshick

Differential Revision: D6793771

fbshipit-source-id: 4716102f41f12801c6f715b6f7ad7e438591b295

Reviewed By: rbgirshick

Differential Revision: D6793159

fbshipit-source-id: 8d8f12915809238d3dcf8066a94446e0feb4cc46

Reviewed By: rbgirshick

Differential Revision: D6784702

fbshipit-source-id: b5f81d858f9e9d33c048c39f849d0e29e8db3cb1