Use different configs for proposal train/test (#587)

* use different configs for proposal train/test

* update configs for dcn

* update inf speed

* update inf speed in dcn, gn, htc

* update comparision

* keep backward compatibility

configs/faster_rcnn_ohem_r50_fpn_1x.py

@@ -58,6 +58,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -76,9 +83,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/faster_rcnn_r101_fpn_1x.py

@@ -58,6 +58,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -76,9 +83,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/faster_rcnn_r50_c4_1x.py

@@ -65,6 +65,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=12000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -83,9 +90,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=12000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=6000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/faster_rcnn_r50_fpn_1x.py

@@ -58,6 +58,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -76,9 +83,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/faster_rcnn_x101_32x4d_fpn_1x.py

@@ -60,6 +60,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -78,9 +85,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/faster_rcnn_x101_64x4d_fpn_1x.py

@@ -60,6 +60,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -78,9 +85,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py

@@ -68,6 +68,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -86,9 +93,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py

@@ -81,6 +81,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -100,9 +107,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py

@@ -81,6 +81,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -100,9 +107,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py

@@ -83,6 +83,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -102,9 +109,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn/README.md

@@ -15,13 +15,14 @@
 
 | Backbone      | model      | Lr schd | Mem (GB) | Train time (s/iter) | Inf time (fps) | box AP | mask AP | Download |
 |:-------------:|:----------:|:-------:|:--------:|:-------------------:|:--------------:|:------:|:-------:|:--------:|
-| R-50-FPN (d)  | Mask R-CNN | 2x      | 7.2      | 0.806               | 5.4            | 39.9   | 36.1    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_2x_20180113-86832cf2.pth) |
-| R-50-FPN (d)  | Mask R-CNN | 3x      | 7.2      | 0.806               | 5.4            | 40.2   | 36.5    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_3x_20180113-8e82f48d.pth) |
-| R-101-FPN (d) | Mask R-CNN | 2x      | 9.9      | 0.970               | 4.8            | 41.6   | 37.1    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r101_fpn_gn_2x_20180113-9598649c.pth) |
-| R-101-FPN (d) | Mask R-CNN | 3x      | 9.9      | 0.970               | 4.8            | 41.7   | 37.3    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r101_fpn_gn_3x_20180113-a14ffb96.pth) |
+| R-50-FPN (d)  | Mask R-CNN | 2x      | 7.2      | 0.806               | 5.4            | 39.8   | 36.1    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_2x_20180113-86832cf2.pth) |
+| R-50-FPN (d)  | Mask R-CNN | 3x      | 7.2      | 0.806               | 5.4            | 40.1   | 36.4    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_3x_20180113-8e82f48d.pth) |
+| R-101-FPN (d) | Mask R-CNN | 2x      | 9.9      | 0.970               | 4.8            | 41.5   | 37.0    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r101_fpn_gn_2x_20180113-9598649c.pth) |
+| R-101-FPN (d) | Mask R-CNN | 3x      | 9.9      | 0.970               | 4.8            | 41.6   | 37.3    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r101_fpn_gn_3x_20180113-a14ffb96.pth) |
 | R-50-FPN (c)  | Mask R-CNN | 2x      | 7.2      | 0.806               | 5.4            | 39.7   | 35.9    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_contrib_2x_20180113-ec93305c.pth) |
-| R-50-FPN (c)  | Mask R-CNN | 3x      | 7.2      | 0.806               | 5.4            | 40.1   | 36.2    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_contrib_3x_20180113-9d230cab.pth) |
+| R-50-FPN (c)  | Mask R-CNN | 3x      | 7.2      | 0.806               | 5.4            | 40.0   | 36.2    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/gn/mask_rcnn_r50_fpn_gn_contrib_3x_20180113-9d230cab.pth) |
 
 **Notes:**
 - (d) means pretrained model converted from Detectron, and (c) means the contributed model pretrained by [@thangvubk](https://github.com/thangvubk).
-- The `3x` schedule is epoch [28, 34, 36].
\ No newline at end of file
+- The `3x` schedule is epoch [28, 34, 36].
+- **Memory, Train/Inf time is outdated.**
\ No newline at end of file

configs/gn/mask_rcnn_r101_fpn_gn_2x.py

@@ -78,6 +78,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -97,9 +104,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn/mask_rcnn_r50_fpn_gn_2x.py

@@ -78,6 +78,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -97,9 +104,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/gn/mask_rcnn_r50_fpn_gn_contrib_2x.py

@@ -78,6 +78,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=dict(
         assigner=dict(
             type='MaxIoUAssigner',
@@ -97,9 +104,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/htc/README.md

@@ -38,11 +38,11 @@ The results on COCO 2017val is shown in the below table. (results on test-dev ar
 
 | Backbone  | Style   | Lr schd | Mem (GB) | Train time (s/iter) | Inf time (fps) | box AP | mask AP | Download |
 |:---------:|:-------:|:-------:|:--------:|:-------------------:|:--------------:|:------:|:-------:|:--------:|
-| R-50-FPN  | pytorch | 1x      | 7.4      | 0.936               | 3.5            | 42.2   | 37.3    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r50_fpn_1x_20190408-878c1712.pth) |
-| R-50-FPN  | pytorch | 20e     | -        | -                   | -              | 43.2   | 38.0    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r50_fpn_20e_20190408-c03b7015.pth) |
-| R-101-FPN | pytorch | 20e     | 9.3      | 1.051               | 3.4            | 44.9   | 39.4    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r101_fpn_20e_20190408-a2e586db.pth) |
-| X-101-32x4d-FPN | pytorch |20e| 5.8      | 0.769               | 3.3            | 46.1   | 40.3    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_x101_32x4d_fpn_20e_20190408-9eae4d0b.pth) |
-| X-101-64x4d-FPN | pytorch |20e| 7.5      | 1.120               | 3.0            | 47.0   | 40.9    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_x101_64x4d_fpn_20e_20190408-497f2561.pth) |
+| R-50-FPN  | pytorch | 1x      | 7.4      | 0.936               | 4.1            | 42.1   | 37.3    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r50_fpn_1x_20190408-878c1712.pth) |
+| R-50-FPN  | pytorch | 20e     | -        | -                   | -              | 43.2   | 38.1    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r50_fpn_20e_20190408-c03b7015.pth) |
+| R-101-FPN | pytorch | 20e     | 9.3      | 1.051               | 4.0            | 44.9   | 39.4    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_r101_fpn_20e_20190408-a2e586db.pth) |
+| X-101-32x4d-FPN | pytorch |20e| 5.8      | 0.769               | 3.8            | 46.1   | 40.3    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_x101_32x4d_fpn_20e_20190408-9eae4d0b.pth) |
+| X-101-64x4d-FPN | pytorch |20e| 7.5      | 1.120               | 3.5            | 46.9   | 40.8    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_x101_64x4d_fpn_20e_20190408-497f2561.pth) |
 
 - In the HTC paper and COCO 2018 Challenge, `score_thr` is set to 0.001 for both baselines and HTC.
 - We use 8 GPUs with 2 images/GPU for R-50 and R-101 models, and 16 GPUs with 1 image/GPU for X-101 models.

configs/htc/htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e.py

@@ -117,6 +117,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=[
         dict(
             assigner=dict(
@@ -171,9 +178,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/htc/htc_r101_fpn_20e.py

@@ -109,6 +109,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=[
         dict(
             assigner=dict(
@@ -163,9 +170,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/htc/htc_r50_fpn_1x.py

@@ -109,6 +109,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=[
         dict(
             assigner=dict(
@@ -163,9 +170,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/htc/htc_r50_fpn_20e.py

@@ -109,6 +109,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=[
         dict(
             assigner=dict(
@@ -163,9 +170,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(

configs/htc/htc_without_semantic_r50_fpn_1x.py

@@ -94,6 +94,13 @@ train_cfg = dict(
         pos_weight=-1,
         smoothl1_beta=1 / 9.0,
         debug=False),
+    rpn_proposal=dict(
+        nms_across_levels=False,
+        nms_pre=2000,
+        nms_post=2000,
+        max_num=2000,
+        nms_thr=0.7,
+        min_bbox_size=0),
     rcnn=[
         dict(
             assigner=dict(
@@ -148,9 +155,9 @@ train_cfg = dict(
 test_cfg = dict(
     rpn=dict(
         nms_across_levels=False,
-        nms_pre=2000,
-        nms_post=2000,
-        max_num=2000,
+        nms_pre=1000,
+        nms_post=1000,
+        max_num=1000,
         nms_thr=0.7,
         min_bbox_size=0),
     rcnn=dict(