metrics for pre-trained FPN RCNN (#693)

* number of gpus * syncbn * pylint * resnetv1c * merge * indent * unitest * trigger * norm args * indent * resnet v1d +0.5% * style * update docs * fix args * trigger build * add test * resolve conflict * Add FPN model * Add FPN train scripts * Fix FPN error, Stay tuned Training on VOC is still going on, I will report the result and log later. * Revert "Sync from dmlc/master" * Revert "Revert "Sync from dmlc/master"" * Update gluoncv/model_zoo/fpn/fpn.py * Fix `FPN` Bugs mAP on VOC07 is 58%, stay tuned. * add faster_rcnn_fpn_resnet50_v1b model * Update gluoncv/model_zoo/fpn/fpn.py * Update gluoncv/model_zoo/fpn/fpn.py * Create Readme.md * Update Readme.md * Update Readme.md * Update Readme.md * Update Readme.md * Update Readme.md * Rename Readme.md to README.md * Update Train and Eval script, Support Eval VOC12 Test. * Update scripts/detection/fpn/eval_fpn_voc12.py * Update scripts/detection/fpn/eval_fpn_voc12.py * Update README.md * Update README.md * Update fpn.py * Update gluoncv/model_zoo/model_zoo.py * Update gluoncv/model_zoo/model_zoo.py * default not to use static alloc to save memory, speed is not significantly impacted. added dilated faster_rcnn_resnet50_v1b added mask_rcnn_resnet101_v1b * fix missing args * small fix * docs * rm unneeded file * rm debug log * Faster RCNN with FPN * rm unnecessary files pylint rm Non-ASCII fix syntax lint rm from .fpn import * stride => strides rm syncbn in rpn rm syncbn arg mask rcnn arg fix missing "s" * rm 's' in anchor_generators * old model compatibility fix * not using RPNHead to keep backward compatibility with old models * _strides * mask rcnn compatibility * docs * rm dilated faster rcnn * mask rcnn w/ fpn * rm undefined functions * change default roi mode to 'align' * trigger build * change name of the fpn networks * model store update * Fix typo (#622) * Improve custo coco compatible detection dataset (#624) * coco det improve for custom datasets * allow flexible image path parser * fix pycocotools _isArrayLike * better comment * clean * Add assertions for invalid class names for VOCDetection (#614) * Add assertions for invalid class names * Add assertions for invalid class names (revision1) * Add assertions/warnings for invalid class names (revision2) * Add assertions/warnings for invalid class names (revision3) * Add assertions/warnings for invalid class names (revision4) * add detection paper (#628) * add bibtex * rephrase * update bibtex * Update PSP Params (#629) * update psp params * update with pin-device_id (#630) * sync bn faster rcnn * pylint * change roi from 7 to 14, since the last fpn model we trained use 14 * add pretrained faster rcnn fpn bn * Update model_zoo.py * Update model_zoo.py * fpn RCNN docs
dmlc · Mar 19, 2019 · 8c022b5 · 8c022b5
1 parent a83ef8c
commit 8c022b5
Show file tree

Hide file tree

Showing 3 changed files with 37 additions and 14 deletions.
diff --git a/docs/model_zoo/detection.rst b/docs/model_zoo/detection.rst
@@ -168,13 +168,19 @@ Checkout Faster-RCNN demo tutorial here: :ref:`sphx_glr_build_examples_detection
 .. table::
    :widths: 50 5 25 20
 
-   +------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
-   | Model                              | Box AP          | Training Command                                                                                                                  | Training Log                                                                                                                     |
-   +====================================+=================+===================================================================================================================================+==================================================================================================================================+
-   | faster_rcnn_resnet50_v1b_coco [2]_ | 37.0/57.8/39.6  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet50_v1b_coco.sh>`_  | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet50_v1b_coco_train.log>`_   |
-   +------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
-   | faster_rcnn_resnet101_v1d_coco [2]_| 40.1/60.9/43.3  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco.sh>`_ | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco_train.log>`_  |
-   +------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
+   +-------------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
+   | Model                                     | Box AP          | Training Command                                                                                                                        | Training Log                                                                                                                     |
+   +===========================================+=================+=========================================================================================================================================+==================================================================================================================================+
+   | faster_rcnn_resnet50_v1b_coco [2]_        | 37.0/57.8/39.6  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet50_v1b_coco.sh>`_        | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet50_v1b_coco_train.log>`_   |
+   +-------------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
+   | faster_rcnn_resnet101_v1d_coco [2]_       | 40.1/60.9/43.3  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco.sh>`_       | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco_train.log>`_  |
+   +-------------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
+   | faster_rcnn_fpn_resnet50_v1b_coco [4]_    | 38.4/60.3/41.4  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_fpn_resnet50_v1b_coco.sh>`_    | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco_train.log>`_  |
+   +-------------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
+   | faster_rcnn_fpn_resnet101_v1d_coco [4]_   | 41.2/62.7/44.8  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_fpn_resnet101_v1d_coco.sh>`_   | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco_train.log>`_  |
+   +-------------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
+   | faster_rcnn_fpn_bn_resnet50_v1b_coco [5]_ | 39.3/61.3/42.9  | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_fpn_bn_resnet50_v1b_coco.sh>`_ | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/detection/faster_rcnn_resnet101_v1d_coco_train.log>`_  |
+   +-------------------------------------------+-----------------+-----------------------------------------------------------------------------------------------------------------------------------------+----------------------------------------------------------------------------------------------------------------------------------+
 
 YOLO-v3
 -------
@@ -213,3 +219,9 @@ Checkout YOLO demo tutorial here: :ref:`sphx_glr_build_examples_detection_demo_y
 .. [3] Redmon, Joseph, and Ali Farhadi. \
        "Yolov3: An incremental improvement." \
        arXiv preprint arXiv:1804.02767 (2018).
+.. [4] Tsung-Yi Lin, Piotr Dollár, Ross Girshick, Kaiming He, Bharath Hariharan, Serge Belongie. \
+       "Feature Pyramid Networks for Object Detection." \
+       IEEE Conference on Computer Vision and Pattern Recognition 2017.
+.. [5] Kaiming He, Ross Girshick, Piotr Dollár. \
+       "Rethinking ImageNet Pre-training." \
+       arXiv preprint arXiv:1811.08883 (2018).
diff --git a/docs/model_zoo/segmentation.rst b/docs/model_zoo/segmentation.rst
@@ -104,11 +104,21 @@ Table of pre-trained models for instance segmentation and their performance.
 
   For instance segmentation task, both box overlap and segmentation overlap based AP are evaluated and reported.
 
-+------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
-| Model                        | Box AP                    | Segm AP                  | Command                                                                                                                            | Training Log                                                                                                                         |
-+==============================+===========================+==========================+====================================================================================================================================+======================================================================================================================================+
-| mask_rcnn_resnet50_v1b_coco  | 38.3/58.7/41.4            | 33.1/54.8/35.0           | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_resnet50_v1b_coco.sh>`_      | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_resnet50_v1b_coco_train.log>`_          |
-+------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
+
+MS COCO
+-------
+
++------------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
+| Model                              | Box AP                    | Segm AP                  | Command                                                                                                                            | Training Log                                                                                                                         |
++====================================+===========================+==========================+====================================================================================================================================+======================================================================================================================================+
+| mask_rcnn_resnet50_v1b_coco        | 38.3/58.7/41.4            | 33.1/54.8/35.0           | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_resnet50_v1b_coco.sh>`_      | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_resnet50_v1b_coco_train.log>`_          |
++------------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
+| mask_rcnn_fpn_resnet50_v1b_coco    | 39.2/61.2/42.2            | 35.4/57.5/37.3           | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_fpn_resnet50_v1b_coco.sh>`_  | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_fpn_resnet50_v1b_coco_train.log>`_      |
++------------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
+| mask_rcnn_resnet101_v1d_coco       | 41.3/61.7/44.4            | 35.2/57.8/36.9           | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_resnet101_v1d_coco.sh>`_     | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_resnet101_v1d_coco_train.log>`_         |
++------------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
+| mask_rcnn_fpn_resnet101_v1d_coco   | 42.3/63.9/46.2            | 37.7/60.5/40.0           | `shell script <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_fpn_resnet101_v1d_coco.sh>`_ | `log <https://raw.githubusercontent.com/dmlc/web-data/master/gluoncv/logs/instance/mask_rcnn_fpn_resnet101_v1d_coco_train.log>`_     |
++------------------------------------+---------------------------+--------------------------+------------------------------------------------------------------------------------------------------------------------------------+--------------------------------------------------------------------------------------------------------------------------------------+
 
 .. [1] He, Kaming, Georgia Gkioxari, Piotr Dollár and Ross Girshick. \
         "Mask R-CNN." \

diff --git a/gluoncv/model_zoo/model_zoo.py b/gluoncv/model_zoo/model_zoo.py
@@ -16,6 +16,7 @@
 from .nasnet import *
 from .pruned_resnet.resnetv1b_pruned import *
 from .pspnet import *
+from .quantized import *
 from .residual_attentionnet import *
 from .resnet import *
 from .resnetv1b import *
@@ -26,7 +27,6 @@
 from .ssd import *
 from .vgg import *
 from .yolo import *
-from .quantized import *
 
 __all__ = ['get_model', 'get_model_list']
 
@@ -198,7 +198,8 @@
     'ssd_512_mobilenet1.0_voc_int8': ssd_512_mobilenet1_0_voc_int8,
     'ssd_512_resnet50_v1_voc_int8': ssd_512_resnet50_v1_voc_int8,
     'ssd_512_vgg16_atrous_voc_int8': ssd_512_vgg16_atrous_voc_int8,
-    }
+}
+
 
 def get_model(name, **kwargs):
     """Returns a pre-defined model by name