Python net_utils._smooth_l1_loss方法代碼示例

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def detect_loss(self, cls_score, rois_label, bbox_pred, rois_target, rois_inside_ws, rois_outside_ws):
        # classification loss
        RCNN_loss_cls = F.cross_entropy(cls_score, rois_label)

        # bounding box regression L1 loss
        RCNN_loss_bbox = _smooth_l1_loss(bbox_pred, rois_target, rois_inside_ws, rois_outside_ws)

        return RCNN_loss_cls, RCNN_loss_bbox 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):
        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, dim=1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'
        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None
            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes[:,:,:5], im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):
        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, 1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                  im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target(
                (rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3,
                                                          1).contiguous().view(
                batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1, 2), 0,
                                               rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)

            rpn_bbox_targets, rpn_bbox_inside_weights, \
            rpn_bbox_outside_weights = rpn_data[
                                       1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets,
                                                rpn_bbox_inside_weights,
                                                rpn_bbox_outside_weights,
                                                sigma=3, dim=[1, 2, 3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:-1]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        if self.training:
            return rois, self.rpn_loss_cls, self.rpn_loss_box, rpn_data[-1]
        return rois, self.rpn_loss_cls, self.rpn_loss_box, None 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, dim=1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):
        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, dim=1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                  im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1, 2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                rpn_bbox_outside_weights, sigma=3, dim=[1, 2, 3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, 1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, 1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)


        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0


        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))

            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())

            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])
        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes):
        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, dim=1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                  im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0
        rpn_label = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1, 2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                rpn_bbox_outside_weights, sigma=3, dim=[1, 2, 3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box, rpn_label, rpn_conv1, rpn_cls_score 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes, need_backprop=None):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        # print(rpn_cls_score_reshape.size())
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape,1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)
        # print(rpn_bbox_pred.max())

        # proposal layer
        # cfg_key = 'TRAIN' if self.training else 'TEST'
        cfg_key = 'TRAIN' if need_backprop.numpy() else 'TEST'
        # cfg_key = 'TRAIN' if need_backprop.numpy() else 'TEST'
        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes, target=False):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, 1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training and not target else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key))

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training and not target:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box 

# 需要導入模塊: from model.utils import net_utils [as 別名]
# 或者: from model.utils.net_utils import _smooth_l1_loss [as 別名]
def forward(self, base_feat, im_info, gt_boxes, num_boxes,target=False):

        batch_size = base_feat.size(0)

        # return feature map after convrelu layer
        rpn_conv1 = F.relu(self.RPN_Conv(base_feat), inplace=True)
        # get rpn classification score
        rpn_cls_score = self.RPN_cls_score(rpn_conv1)

        rpn_cls_score_reshape = self.reshape(rpn_cls_score, 2)
        rpn_cls_prob_reshape = F.softmax(rpn_cls_score_reshape, 1)
        rpn_cls_prob = self.reshape(rpn_cls_prob_reshape, self.nc_score_out)

        # get rpn offsets to the anchor boxes
        rpn_bbox_pred = self.RPN_bbox_pred(rpn_conv1)

        # proposal layer
        cfg_key = 'TRAIN' if self.training else 'TEST'

        rois = self.RPN_proposal((rpn_cls_prob.data, rpn_bbox_pred.data,
                                 im_info, cfg_key),target=target)

        self.rpn_loss_cls = 0
        self.rpn_loss_box = 0

        # generating training labels and build the rpn loss
        if self.training:
            assert gt_boxes is not None

            rpn_data = self.RPN_anchor_target((rpn_cls_score.data, gt_boxes, im_info, num_boxes))

            # compute classification loss
            rpn_cls_score = rpn_cls_score_reshape.permute(0, 2, 3, 1).contiguous().view(batch_size, -1, 2)
            rpn_label = rpn_data[0].view(batch_size, -1)

            rpn_keep = Variable(rpn_label.view(-1).ne(-1).nonzero().view(-1))
            rpn_cls_score = torch.index_select(rpn_cls_score.view(-1,2), 0, rpn_keep)
            rpn_label = torch.index_select(rpn_label.view(-1), 0, rpn_keep.data)
            rpn_label = Variable(rpn_label.long())
            self.rpn_loss_cls = F.cross_entropy(rpn_cls_score, rpn_label)
            fg_cnt = torch.sum(rpn_label.data.ne(0))

            rpn_bbox_targets, rpn_bbox_inside_weights, rpn_bbox_outside_weights = rpn_data[1:]

            # compute bbox regression loss
            rpn_bbox_inside_weights = Variable(rpn_bbox_inside_weights)
            rpn_bbox_outside_weights = Variable(rpn_bbox_outside_weights)
            rpn_bbox_targets = Variable(rpn_bbox_targets)

            self.rpn_loss_box = _smooth_l1_loss(rpn_bbox_pred, rpn_bbox_targets, rpn_bbox_inside_weights,
                                                            rpn_bbox_outside_weights, sigma=3, dim=[1,2,3])

        return rois, self.rpn_loss_cls, self.rpn_loss_box