1d7b7345ab3254ed9b450dbd4208a3d63cb4963f,dataset/models/tf/fcn.py,FCN,_build,#FCN#,20

Before Change

        layers_dicts = {"conv": conv, "batch_norm": batch_norm}

        net = VGG16.body(dim, inputs["images"], b_norm, **layers_dicts)
        net = conv_block(dim, net, 100, 7, "ca", "conv-out-1", **layers_dicts)
        net = conv_block(dim, net, 100, 1, "ca", "conv-out-2", padding="VALID", **layers_dicts)
        net = conv_block(dim, net, n_classes, 1, "ca", "conv-out-3", padding="VALID", **layers_dicts)
        conv7 = net
        pool4 = tf.get_default_graph().get_tensor_by_name("body/block-3/output:0")
        pool3 = tf.get_default_graph().get_tensor_by_name("body/block-2/output:0")
        if arch == "FCN32":

After Change

        layers_dicts = {"conv": conv, "batch_norm": batch_norm}

        net = VGG16.body(dim, inputs["images"], enable_batch_norm, **layers_dicts)
        layout = "cna" * 3 if enable_batch_norm else "ca" * 3
        net = conv_block(dim, net, [100, 100, n_classes], [7, 1, 1], layout, "conv-out", 
                         padding=["SAME", "VALID", "VALID"], **layers_dicts)
        conv7 = net
        pool4 = tf.get_default_graph().get_tensor_by_name("body/block-3/output:0")

Instances