Method __init__

network/ocr_utils.py:128–147

(self, in_channels, key_channels, out_channels, scale=1,
                 dropout=0.1)

    for each pixel.
    """
    def __init__(self, in_channels, key_channels, out_channels, scale=1,
                 dropout=0.1):
        super(SpatialOCR_Module, self).__init__()
        self.object_context_block = ObjectAttentionBlock(in_channels,
                                                         key_channels,
                                                         scale)
        if cfg.MODEL.OCR_ASPP:
            self.aspp, aspp_out_ch = get_aspp(
                in_channels, bottleneck_ch=cfg.MODEL.ASPP_BOT_CH,
                output_stride=8)
            _in_channels = 2 * in_channels + aspp_out_ch
        else:
            _in_channels = 2 * in_channels

        self.conv_bn_dropout = nn.Sequential(
            nn.Conv2d(_in_channels, out_channels, kernel_size=1, padding=0,
                      bias=False),
            BNReLU(out_channels),
            nn.Dropout2d(dropout)
        )

    def forward(self, feats, proxy_feats):
        context = self.object_context_block(feats, proxy_feats)