LAP/torchlap/utils/random_brightness_augmentation.py

from random import random

import torch


class ClippedBrightnessAugment(torch.nn.Module):
    def __init__(self, min_factor, max_factor, clip_min=-1 * float('inf'), clip_max=float('inf')):
        super().__init__()
        self._min_factor = min_factor
        self._max_factor = max_factor
        self._clip_min = clip_min
        self._clip_max = clip_max

    def forward(self, x):
        random_scale = random()
        random_scale = random_scale * (self._max_factor - self._min_factor) + self._min_factor
        return torch.clip(x * random_scale, self._clip_min, self._clip_max)