reproduce box-1x

examples/FasterRCNN/config.py

@@ -25,7 +25,7 @@ WARMUP = 1000    # in steps
 STEPS_PER_EPOCH = 500
 LR_SCHEDULE = [150000, 230000, 280000]
 LR_SCHEDULE = [120000, 160000, 180000]    # "1x" schedule in detectron
-#LR_SCHEDULE = [240000, 320000, 360000]    # "2x" schedule in detectron
+LR_SCHEDULE = [240000, 320000, 360000]    # "2x" schedule in detectron
 
 # image resolution --------------------
 SHORT_EDGE_SIZE = 800

examples/FasterRCNN/data.py

@@ -144,14 +144,16 @@ def get_anchor_labels(anchors, gt_boxes, crowd_boxes):
     # Subsample fg labels: ignore some fg if fg is too many
     target_num_fg = int(config.RPN_BATCH_PER_IM * config.RPN_FG_RATIO)
     fg_inds = filter_box_label(anchor_labels, 1, target_num_fg)
+    if len(fg_inds) == 0:
+        raise MalformedData("No valid foreground for RPN!")
     # Note that fg could be fewer than the target ratio
 
     # Subsample bg labels. num_bg is not allowed to be too many
     old_num_bg = np.sum(anchor_labels == 0)
-    if old_num_bg == 0 or len(fg_inds) == 0:
+    if old_num_bg == 0:
         # No valid bg/fg in this image, skip.
         # This can happen if, e.g. the image has large crowd.
-        raise MalformedData("No valid foreground/background for RPN!")
+        raise MalformedData("No valid background for RPN!")
     target_num_bg = config.RPN_BATCH_PER_IM - len(fg_inds)
     filter_box_label(anchor_labels, 0, target_num_bg)   # ignore return values
 
@@ -341,7 +343,7 @@ def get_train_dataflow(add_mask=False):
             # tpviz.interactive_imshow(viz)
         return ret
 
-    ds = MultiProcessMapDataZMQ(ds, 5, preprocess)
+    ds = MultiProcessMapDataZMQ(ds, 10, preprocess)
     #ds = PrefetchDataZMQ(ds, 3)
     return ds
 

examples/FasterRCNN/model.py

@@ -10,6 +10,7 @@ from tensorpack.models import (
     Conv2D, FullyConnected, MaxPooling,
     layer_register, Conv2DTranspose, FixedUnPooling)
 
+from tensorpack.utils import logger
 from utils.box_ops import pairwise_iou
 from utils.box_ops import area as tf_area
 import config
@@ -578,7 +579,13 @@ def fpn_model(features):
 
     def upsample2x(name, x):
         # TODO may not be optimal in speed or math
+        logger.info("Unpool 1111 ...")
+        return FixedUnPooling(
+            name, x, 2, unpool_mat=np.ones((2, 2), dtype='float32'),
+            data_format='channels_first')
+
         with tf.name_scope(name):
+            logger.info("Nearest neighbor")
             shape2d = tf.shape(x)[2:]
             x = tf.transpose(x, [0, 2, 3, 1])
             x = tf.image.resize_nearest_neighbor(x, shape2d * 2, align_corners=True)
@@ -616,7 +623,8 @@ def fpn_map_rois_to_levels(boxes):
     Be careful that the returned tensor could be empty.
     """
     sqrtarea = tf.sqrt(tf_area(boxes))
-    level = tf.floor(4 + tf.log(sqrtarea * (1. / 224) + 1e-6) * (1.0 / np.log(2)))
+    level = tf.to_int32(tf.floor(
+        4 + tf.log(sqrtarea * (1. / 224) + 1e-6) * (1.0 / np.log(2))))
 
     # RoI levels range from 2~5 (not 6)
     level_ids = [
@@ -645,6 +653,7 @@ def fastrcnn_2fc_head(feature, dim, num_classes):
     Returns:
         cls_logits (Nxnum_class), reg_logits (Nx num_class-1 x 4)
     """
+    logger.info("fc-head-stddev=0.01")
     init = tf.random_normal_initializer(stddev=0.01)
     hidden = FullyConnected('fc6', feature, dim, kernel_initializer=init, nl=tf.nn.relu)
     hidden = FullyConnected('fc7', hidden, dim, kernel_initializer=init, nl=tf.nn.relu)