ProGamerGov

neural-style-pt with mean loss and histogram transfer

Histogram Transfer

Users can specify an image for which the histogram will be transfered from, and what images the histogram will be transfered to; either the content image, style image(s), or both.

Mean Loss

A new loss layer type has been added that uses image means. Currently it only uses the first style image specified.

ProGamerGov

import os
import copy
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms

from PIL import Image
from CaffeLoader import loadCaffemodel, ModelParallel

ProGamerGov

#!/bin/bash
fileid="1qWjJd-mVNxvHhzc9XsU6LjC0pQXIwW4t"
filename="models/VGG16-Stylized-ImageNet.pth"
curl -c ./cookie -s -L "https://drive.google.com/uc?export=download&id=${fileid}" > /dev/null
curl -Lb ./cookie "https://drive.google.com/uc?export=download&confirm=`awk '/download/ {print $NF}' ./cookie`&id=${fileid}" -o ${filename}

fileid="1EU-F9ugeIeTO9ay4PinzsBXgEuCYBu0Z"
filename="models/VGG16_SOD_finetune.pth"
curl -c ./cookie -s -L "https://drive.google.com/uc?export=download&id=${fileid}" > /dev/null
curl -Lb ./cookie "https://drive.google.com/uc?export=download&confirm=`awk '/download/ {print $NF}' ./cookie`&id=${fileid}" -o ${filename}

ProGamerGov

import torch
from collections import OrderedDict

import argparse
parser = argparse.ArgumentParser()
parser.add_argument("-model_file", type=str, default='')
parser.add_argument("-output_name", type=str, default='')
params = parser.parse_args()

ProGamerGov

ProGamerGov

import os
import copy
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms

from PIL import Image
from CaffeLoader import loadCaffemodel, ModelParallel

ProGamerGov

import os
import copy
import time
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms

from PIL import Image
from CaffeLoader import loadCaffemodel, ModelParallel

ProGamerGov

import os
import copy
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.transforms as transforms

from PIL import Image
from CaffeLoader import loadCaffemodel, ModelParallel

ProGamerGov

import torch
import torch.nn as nn


class VGG(nn.Module):
    def __init__(self, features, num_classes=1000):
        super(VGG, self).__init__()
        self.features = features
        self.classifier = nn.Sequential(
            nn.Linear(512 * 7 * 7, 4096),

ProGamerGov

# Inspired by: https://github.com/torch/nn/blob/master/GPU.lua
# And: https://github.com/jcjohnson/neural-style/blob/master/neural_style.lua#L360
# As seen in: https://github.com/ProGamerGov/neural-style-pt

import torch
import torch.nn as nn


class ModelParallel(nn.Module):
    r"""Splits a sequential network across multiple devices.