arcface_loss.py

import numpy as np
import torch

from ..utils import common_functions as c_f
from .large_margin_softmax_loss import LargeMarginSoftmaxLoss


class ArcFaceLoss(LargeMarginSoftmaxLoss):
    """
    Implementation of https://arxiv.org/pdf/1801.07698.pdf
    """

    def __init__(self, num_classes, embedding_size, margin=28.6, scale=64, **kwargs):
        super().__init__(
            num_classes=num_classes,
            embedding_size=embedding_size,
            margin=margin,
            scale=scale,
            **kwargs
        )

    def init_margin(self):
        self.margin = np.radians(self.margin)

    def cast_types(self, dtype, device):
        self.W.data = c_f.to_device(self.W.data, device=device, dtype=dtype)

    def modify_cosine_of_target_classes(self, cosine_of_target_classes):
        angles = self.get_angles(cosine_of_target_classes)

        # Compute cos of (theta + margin) and cos of theta
        cos_theta_plus_margin = torch.cos(angles + self.margin)
        cos_theta = torch.cos(angles)

        # Keep the cost function monotonically decreasing
        unscaled_logits = torch.where(
            angles <= np.deg2rad(180) - self.margin,
            cos_theta_plus_margin,
            cos_theta - self.margin * np.sin(self.margin),
        )

        return unscaled_logits

    def scale_logits(self, logits, *_):
        return logits * self.scale