def __init__(self, **kwags):
super(RPN, self).__init__(**kwags)
# Anchor attributes
self.ANCHOR_SCALES = (32, 64, 128, 256, 512)
self.ANCHOR_RATIOS = (0.5, 1, 2)
# The strides of each layer of the FPN Pyramid.
self.FEATURE_STRIDES = (4, 8, 16, 32, 64)
# Bounding box refinement mean and standard deviation
self.RPN_TARGET_MEANS = (0., 0., 0., 0.)
self.RPN_TARGET_STDS = (0.1, 0.1, 0.2, 0.2)
self.backbone = resnet.ResNet(depth=101, name='res_net')
self.neck = fpn.FPN(name='fpn')
self.rpn_head = rpn_head.RPNHead(anchors_per_location=len(
self.ANCHOR_RATIOS),
name='rpn_head')
self.generator = anchor_generator.AnchorGenerator(
scales=self.ANCHOR_SCALES,
ratios=self.ANCHOR_RATIOS,
feature_strides=self.FEATURE_STRIDES)
self.anchor_target = anchor_target.AnchorTarget(
target_means=self.RPN_TARGET_MEANS,
target_stds=self.RPN_TARGET_STDS)
self.rpn_class_loss = losses.rpn_class_loss
self.rpn_bbox_loss = losses.rpn_bbox_loss
def __init__(self,
anchor_scales=(32, 64, 128, 256, 512),
anchor_ratios=(0.5, 1, 2),
anchor_feature_strides=(4, 8, 16, 32, 64),
proposal_count=2000,
nms_threshold=0.7,
target_means=(0., 0., 0., 0.),
target_stds=(0.1, 0.1, 0.2, 0.2),
num_rpn_deltas=256,
positive_fraction=0.5,
pos_iou_thr=0.7,
neg_iou_thr=0.3,
**kwags):
'''
Network head of Region Proposal Network.
/ - rpn_cls (1x1 conv)
input - rpn_conv (3x3 conv) -
\ - rpn_reg (1x1 conv)
Attributes
---
anchor_scales: 1D array of anchor sizes in pixels.
anchor_ratios: 1D array of anchor ratios of width/height.
anchor_feature_strides: Stride of the feature map relative
to the image in pixels.
proposal_count: int. RPN proposals kept after non-maximum
suppression.
nms_threshold: float. Non-maximum suppression threshold to
filter RPN proposals.
target_means: [4] Bounding box refinement mean.
target_stds: [4] Bounding box refinement standard deviation.
num_rpn_deltas: int.
positive_fraction: float.
pos_iou_thr: float.
neg_iou_thr: float.
'''
super(RPNHead, self).__init__(**kwags)
self.proposal_count = proposal_count
self.nms_threshold = nms_threshold
self.target_means = target_means
self.target_stds = target_stds
self.generator = anchor_generator.AnchorGenerator(
scales=anchor_scales,
ratios=anchor_ratios,
feature_strides=anchor_feature_strides)
self.anchor_target = anchor_target.AnchorTarget(
target_means=target_means,
target_stds=target_stds,
num_rpn_deltas=num_rpn_deltas,
positive_fraction=positive_fraction,
pos_iou_thr=pos_iou_thr,
neg_iou_thr=neg_iou_thr)
self.rpn_class_loss = losses.rpn_class_loss
self.rpn_bbox_loss = losses.rpn_bbox_loss
# Shared convolutional base of the RPN
self.rpn_conv_shared = layers.Conv2D(512, (3, 3), padding='same',
kernel_initializer='he_normal',
name='rpn_conv_shared')
self.rpn_class_raw = layers.Conv2D(len(anchor_ratios) * 2, (1, 1),
kernel_initializer='he_normal',
name='rpn_class_raw')
self.rpn_delta_pred = layers.Conv2D(len(anchor_ratios) * 4, (1, 1),
kernel_initializer='he_normal',
name='rpn_bbox_pred')
def __init__(self, num_classes, **kwags):
super(FasterRCNN, self).__init__(**kwags)
self.NUM_CLASSES = num_classes
# Anchor attributes
self.ANCHOR_SCALES = (32, 64, 128, 256, 512)
self.ANCHOR_RATIOS = (0.5, 1, 2)
# The strides of each layer of the FPN Pyramid.
self.FEATURE_STRIDES = (4, 8, 16, 32, 64)
# Bounding box refinement mean and standard deviation
self.RPN_TARGET_MEANS = (0., 0., 0., 0.)
self.RPN_TARGET_STDS = (0.1, 0.1, 0.2, 0.2)
self.PRN_PROPOSAL_COUNT = 2000
self.PRN_NMS_THRESHOLD = 0.7
self.ROI_BATCH_SIZE = 512
# Bounding box refinement mean and standard deviation
self.RCNN_TARGET_MEANS = (0., 0., 0., 0.)
self.RCNN_TARGET_STDS = (0.1, 0.1, 0.2, 0.2)
self.POOL_SIZE = (7, 7)
self.backbone = resnet.ResNet(depth=101, name='res_net')
self.neck = fpn.FPN(name='fpn')
self.rpn_head = rpn_head.RPNHead(
anchors_per_location=len(self.ANCHOR_RATIOS),
proposal_count=self.PRN_PROPOSAL_COUNT,
nms_threshold=self.PRN_NMS_THRESHOLD,
target_means=self.RPN_TARGET_MEANS,
target_stds=self.RPN_TARGET_STDS,
name='rpn_head')
self.roi_align = roi_align.PyramidROIAlign(pool_shape=self.POOL_SIZE,
name='pyramid_roi_align')
self.bbox_head = bbox_head.BBoxHead(num_classes=self.NUM_CLASSES,
pool_size=self.POOL_SIZE,
name='b_box_head')
self.generator = anchor_generator.AnchorGenerator(
scales=self.ANCHOR_SCALES,
ratios=self.ANCHOR_RATIOS,
feature_strides=self.FEATURE_STRIDES)
self.anchor_target = anchor_target.AnchorTarget(
target_means=self.RPN_TARGET_MEANS,
target_stds=self.RPN_TARGET_STDS)
self.bbox_target = bbox_target.ProposalTarget(
target_means=self.RCNN_TARGET_MEANS,
target_stds=self.RPN_TARGET_STDS,
num_rcnn_deltas=self.ROI_BATCH_SIZE)
self.rpn_class_loss = losses.rpn_class_loss
self.rpn_bbox_loss = losses.rpn_bbox_loss
self.rcnn_class_loss = losses.rcnn_class_loss
self.rcnn_bbox_loss = losses.rcnn_bbox_loss