Handle data that belongs to classes not seen in training or testing

TauvicR · November 5, 2020, 9:03am

Agnostophobia, the fear of the unknown, can be experienced by deep learning
engineers while applying their networks to real-world applications.

The solution may be burried somewhere in here:

TauvicR · November 6, 2020, 8:45am

I think that implementing a Ring loss will help to handle unseen images

github.com

Paralysis/ringloss/blob/master/loss_func.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn.parameter import Parameter
from torch.autograd import Function
import numpy as np

class SoftmaxLoss(nn.Module):
    def __init__(self, input_size, output_size, normalize=True, loss_weight=1.0):
        super(SoftmaxLoss, self).__init__()
        self.fc = nn.Linear(input_size, int(output_size), bias=False)
        nn.init.kaiming_uniform_(self.fc.weight, 0.25)
        self.weight = self.fc.weight
        self.loss_weight = loss_weight
        self.normalize = normalize

    def forward(self, x, y):
        if self.normalize:
            self.fc.weight.renorm(2, 0, 1e-5).mul(1e5)
        prob = F.log_softmax(self.fc(x), dim=1)

This file has been truncated. show original