symbol_factory.py 文件源码

def get_scales(min_scale=0.2, max_scale=0.9,num_layers=6):
    """ Following the ssd arxiv paper, regarding the calculation of scales & ratios

    Parameters
    ----------
    min_scale : float
    max_scales: float
    num_layers: int
        number of layers that will have a detection head
    anchor_ratios: list
    first_layer_ratios: list

    return
    ------
    sizes : list
        list of scale sizes per feature layer
    ratios : list
        list of anchor_ratios per feature layer
    """

    # this code follows the original implementation of wei liu
    # for more, look at ssd/score_ssd_pascal.py:310 in the original caffe implementation
    min_ratio = int(min_scale * 100)
    max_ratio = int(max_scale * 100)
    step = int(np.floor((max_ratio - min_ratio) / (num_layers - 2)))
    min_sizes = []
    max_sizes = []
    for ratio in xrange(min_ratio, max_ratio + 1, step):
        min_sizes.append(ratio / 100.)
        max_sizes.append((ratio + step) / 100.)
    min_sizes = [int(100*min_scale / 2.0) / 100.0] + min_sizes
    max_sizes = [min_scale] + max_sizes

    # convert it back to this implementation's notation:
    scales = []
    for layer_idx in range(num_layers):
        scales.append([min_sizes[layer_idx], np.single(np.sqrt(min_sizes[layer_idx] * max_sizes[layer_idx]))])
    return scales