StarGAN2 のバックアップ(No.11) - PukiWiki

[ トップ ] [ 一覧 | 検索 | 履歴 | ログイン ]

私的AI研究会 > StarGAN2

顔の合成：StarGAN-v2（その２）== 編集中 == †

　「StarGAN-V2」で特徴操作・顔の合成をおこない顔を変化させる
　ローカル環境で「StarGAN-V2」を動かす

顔の合成：StarGAN-v2（その２）== 編集中 ==
参考資料

※ 最終更新:2024/08/06　

「StarGAN-V2」 †

概要 †

StarGAN v2 は韓国の Naver 社 Clova AI Rsearch チームによって提案されたスタイル変換モデル
ベースモデルには GANs の一種である StarGAN と呼ばれるモデルが使用されている

モデル概要図（下記論文所収）
論文「StarGAN v2: Diverse Image Synthesis for Multiple Domains」
<paper>
・https://arxiv.org/pdf/1912.01865
<framework>
・https://github.com/clovaai/stargan-v2

前回の検証
・顔の合成：StarGAN-v2

実行環境の構築 †

仮想環境「py38_learn」で実行する
未作成の場合は → 『仮想環境 (py38_learn)』の手順で仮想環境を作成

GitHub サイトからプロジェクトをダウンロード

cd /anaconda_win/workspace_2　　　　　　　　　　　　　　　　　　　　　← Windows の場合
cd ~/workspace_2　　　　　　　　　　　　　　　　　　　　　　　　　　　← Linux の場合

git clone https://github.com/clovaai/stargan-v2.git

プロジェクト・パッケージ project_stargan-v2.zip (661MB) <stargan-v2> をダウンロード
・解凍してできるフォルダ

update
└─workspace_2
    └─stargan-v2　　　　　　　　　　　　　　　　　　　　　　　　← GitHub からクローンしたプロジェクトに上書きする
        ├─assets
        │  └─representative
        ├─core
        └─expr
            ├─checkpoints
            └─results

・解凍してできる「update/」フォルダ以下を次のフォルダの下に上書きコピーする
　Windows の場合 →「anaconda_win/」　Linux の場合 → 「~/」

提供されているデモ「main.py」を実行 †

コマンド

(py38_gan) python main.py --mode sample --num_domains 2 --resume_iter 100000 --w_hpf 1 --checkpoint_dir expr/checkpoints/celeba_hq --result_dir expr/results/celeba_hq --src_dir assets/representative/celeba_hq/src_0 --ref_dir assets/representative/celeba_hq/ref_0

▼「python main.py --mode sample ... 」ログ

(py38_learn) python main.py --mode sample --num_domains 2 --resume_iter 100000 --w_hpf 1 --checkpoint_dir expr/checkpoints/celeba_hq --result_dir expr/results/celeba_hq --src_dir assets/representative/celeba_hq/src_0 --ref_dir assets/representative/celeba_hq/ref_0
Namespace(batch_size=8, beta1=0.0, beta2=0.99, checkpoint_dir='expr/checkpoints/celeba_hq', ds_iter=100000, eval_dir='expr/eval', eval_every=50000, f_lr=1e-06, hidden_dim=512, img_size=256, inp_dir='assets/representative/custom/female', lambda_cyc=1, lambda_ds=1, lambda_reg=1, lambda_sty=1, latent_dim=16, lm_path='expr/checkpoints/celeba_lm_mean.npz', lr=0.0001, mode='sample', num_domains=2, num_outs_per_domain=10, num_workers=4, out_dir='assets/representative/celeba_hq/src/female', print_every=10, randcrop_prob=0.5, ref_dir='assets/representative/celeba_hq/ref_0', result_dir='expr/results/celeba_hq', resume_iter=100000, sample_dir='expr/samples', sample_every=5000, save_every=10000, seed=777, src_dir='assets/representative/celeba_hq/src_0', style_dim=64, total_iters=100000, train_img_dir='data/celeba_hq/train', val_batch_size=32, val_img_dir='data/celeba_hq/val', w_hpf=1.0, weight_decay=0.0001, wing_path='expr/checkpoints/wing.ckpt')
Number of parameters of generator: 43467395
Number of parameters of mapping_network: 2438272
Number of parameters of style_encoder: 20916928
Number of parameters of discriminator: 20852290
Number of parameters of fan: 6333603
Initializing generator...
Initializing mapping_network...
Initializing style_encoder...
Initializing discriminator...
Preparing DataLoader for the generation phase...
Preparing DataLoader for the generation phase...
Loading checkpoint from expr/checkpoints/celeba_hq\100000_nets_ema.ckpt...
Working on expr/results/celeba_hq\reference.jpg...

実行結果
・オフィシャルサイトのコマンドパラメータを一部変更（実行時間を短縮するため）
　「src」→「src_0」「ref」→「ref_0」
　「src」「ref」フォルダ内（female/,male/）から1画像づつ抽出して「src_0」「ref_0」としている
・終了すれば「stargan-v2/expr/results/celeba_hq/」に画像ファイル「reference.jpg」が作成される

改良版「main2.py」プログラムを実行する

・プログラム終了後にできた画像ファイルを画面に収まるように縮小表示する
・生成画像を表示 3秒後プログラムは終了する

(py38_gan) python main2.py --mode sample --num_domains 2 --resume_iter 100000 --w_hpf 1 --checkpoint_dir expr/checkpoints/celeba_hq --result_dir expr/results/celeba_hq --src_dir assets/representative/celeba_hq/src_0 --ref_dir assets/representative/celeba_hq/ref_0

▼「main2.py」ソースコード

# -*- coding: utf-8 -*-
##------------------------------------------
##  StarGAN v2      Ver 0.02
##    Generating interpolation image
##
##               2023.10.02 Masahiro Izutsu
##------------------------------------------
## Official StarGAN v2 https://github.com/clovaai/stargan-v2
##
## main2.py        (original: main.py)
##  Ver 0.02    2024.07.08  ウインドウ表示「X」ボタン有効

"""
StarGAN v2
Copyright (c) 2020-present NAVER Corp.

This work is licensed under the Creative Commons Attribution-NonCommercial
4.0 International License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc/4.0/ or send a letter to
Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.
"""

import warnings
warnings.simplefilter('ignore')

import os
import argparse

from munch import Munch
from torch.backends import cudnn
import torch

from core.data_loader import get_train_loader
from core.data_loader import get_test_loader
from core.solver import Solver

import my_imagetool                                                             # 2024.07.08

def str2bool(v):
    return v.lower() in ('true')


def subdirs(dname):
    return [d for d in os.listdir(dname)
            if os.path.isdir(os.path.join(dname, d))]


def main(args):
    print(args)
    cudnn.benchmark = True
    torch.manual_seed(args.seed)

    solver = Solver(args)

    if args.mode == 'train':
        assert len(subdirs(args.train_img_dir)) == args.num_domains
        assert len(subdirs(args.val_img_dir)) == args.num_domains
        loaders = Munch(src=get_train_loader(root=args.train_img_dir,
                                             which='source',
                                             img_size=args.img_size,
                                             batch_size=args.batch_size,
                                             prob=args.randcrop_prob,
                                             num_workers=args.num_workers),
                        ref=get_train_loader(root=args.train_img_dir,
                                             which='reference',
                                             img_size=args.img_size,
                                             batch_size=args.batch_size,
                                             prob=args.randcrop_prob,
                                             num_workers=args.num_workers),
                        val=get_test_loader(root=args.val_img_dir,
                                            img_size=args.img_size,
                                            batch_size=args.val_batch_size,
                                            shuffle=True,
                                            num_workers=args.num_workers))
        solver.train(loaders)
    elif args.mode == 'sample':
        assert len(subdirs(args.src_dir)) == args.num_domains
        assert len(subdirs(args.ref_dir)) == args.num_domains
        loaders = Munch(src=get_test_loader(root=args.src_dir,
                                            img_size=args.img_size,
                                            batch_size=args.val_batch_size,
                                            shuffle=False,
                                            num_workers=args.num_workers),
                        ref=get_test_loader(root=args.ref_dir,
                                            img_size=args.img_size,
                                            batch_size=args.val_batch_size,
                                            shuffle=False,
                                            num_workers=args.num_workers))
        solver.sample(loaders)
        my_imagetool.image2disp(args.result_dir + '/reference.jpg')             # 2024.07.08

    elif args.mode == 'eval':
        solver.evaluate()
    elif args.mode == 'align':
        from core.wing import align_faces
        align_faces(args, args.inp_dir, args.out_dir)
    else:
        raise NotImplementedError


if __name__ == '__main__':
    import datetime

    parser = argparse.ArgumentParser()

    # model arguments
    parser.add_argument('--img_size', type=int, default=256,
                        help='Image resolution')
    parser.add_argument('--num_domains', type=int, default=2,
                        help='Number of domains')
    parser.add_argument('--latent_dim', type=int, default=16,
                        help='Latent vector dimension')
    parser.add_argument('--hidden_dim', type=int, default=512,
                        help='Hidden dimension of mapping network')
    parser.add_argument('--style_dim', type=int, default=64,
                        help='Style code dimension')

    # weight for objective functions
    parser.add_argument('--lambda_reg', type=float, default=1,
                        help='Weight for R1 regularization')
    parser.add_argument('--lambda_cyc', type=float, default=1,
                        help='Weight for cyclic consistency loss')
    parser.add_argument('--lambda_sty', type=float, default=1,
                        help='Weight for style reconstruction loss')
    parser.add_argument('--lambda_ds', type=float, default=1,
                        help='Weight for diversity sensitive loss')
    parser.add_argument('--ds_iter', type=int, default=100000,
                        help='Number of iterations to optimize diversity sensitive loss')
    parser.add_argument('--w_hpf', type=float, default=1,
                        help='weight for high-pass filtering')

    # training arguments
    parser.add_argument('--randcrop_prob', type=float, default=0.5,
                        help='Probabilty of using random-resized cropping')
    parser.add_argument('--total_iters', type=int, default=100000,
                        help='Number of total iterations')
    parser.add_argument('--resume_iter', type=int, default=0,
                        help='Iterations to resume training/testing')
    parser.add_argument('--batch_size', type=int, default=8,
                        help='Batch size for training')
    parser.add_argument('--val_batch_size', type=int, default=32,
                        help='Batch size for validation')
    parser.add_argument('--lr', type=float, default=1e-4,
                        help='Learning rate for D, E and G')
    parser.add_argument('--f_lr', type=float, default=1e-6,
                        help='Learning rate for F')
    parser.add_argument('--beta1', type=float, default=0.0,
                        help='Decay rate for 1st moment of Adam')
    parser.add_argument('--beta2', type=float, default=0.99,
                        help='Decay rate for 2nd moment of Adam')
    parser.add_argument('--weight_decay', type=float, default=1e-4,
                        help='Weight decay for optimizer')
    parser.add_argument('--num_outs_per_domain', type=int, default=10,
                        help='Number of generated images per domain during sampling')

    # misc
    parser.add_argument('--mode', type=str, required=True,
                        choices=['train', 'sample', 'eval', 'align'],
                        help='This argument is used in solver')
    parser.add_argument('--num_workers', type=int, default=4,
                        help='Number of workers used in DataLoader')
    parser.add_argument('--seed', type=int, default=777,
                        help='Seed for random number generator')

    # directory for training
    parser.add_argument('--train_img_dir', type=str, default='data/celeba_hq/train',
                        help='Directory containing training images')
    parser.add_argument('--val_img_dir', type=str, default='data/celeba_hq/val',
                        help='Directory containing validation images')
    parser.add_argument('--sample_dir', type=str, default='expr/samples',
                        help='Directory for saving generated images')
    parser.add_argument('--checkpoint_dir', type=str, default='expr/checkpoints',
                        help='Directory for saving network checkpoints')

    # directory for calculating metrics
    parser.add_argument('--eval_dir', type=str, default='expr/eval',
                        help='Directory for saving metrics, i.e., FID and LPIPS')

    # directory for testing
    parser.add_argument('--result_dir', type=str, default='expr/results',
                        help='Directory for saving generated images and videos')
    parser.add_argument('--src_dir', type=str, default='assets/representative/celeba_hq/src',
                        help='Directory containing input source images')
    parser.add_argument('--ref_dir', type=str, default='assets/representative/celeba_hq/ref',
                        help='Directory containing input reference images')
    parser.add_argument('--inp_dir', type=str, default='assets/representative/custom/female',
                        help='input directory when aligning faces')
    parser.add_argument('--out_dir', type=str, default='assets/representative/celeba_hq/src/female',
                        help='output directory when aligning faces')

    # face alignment
    parser.add_argument('--wing_path', type=str, default='expr/checkpoints/wing.ckpt')
    parser.add_argument('--lm_path', type=str, default='expr/checkpoints/celeba_lm_mean.npz')

    # step size
    parser.add_argument('--print_every', type=int, default=10)
    parser.add_argument('--sample_every', type=int, default=5000)
    parser.add_argument('--save_every', type=int, default=10000)
    parser.add_argument('--eval_every', type=int, default=50000)

    args = parser.parse_args()

    start_time = datetime.datetime.now()        # 時間計測開始

    main(args)

    # 経過時間
    end_time = datetime.datetime.now()
    print(start_time.strftime('\nprocessing start >>\t %Y/%m/%d %H:%M:%S'))
    print(end_time.strftime('processing end >>\t %Y/%m/%d %H:%M:%S'))
    print('processing time >>\t', end_time - start_time)

    print('\nFinished.\n')

GUI で操作できるカスタム画像の前処理プログラム「starv2_align.py」を作成 †

主な機能
・変換対象のカスタム画像の入っているフォルダを ダイアログにより選択指定する
　カスタム画像は全体に占める顔の割合がサンプル画像と同じ割合になるよう予めトリミングしておく

出力ファイルの保存場所
・「--out_dir」パラメータで指定したフォルダ内に「 <変換対象フォルダ名> + '_align'」フォルダを作成し保存

コマンドオプション一覧

コマンドオプション	引数	初期値	意味
--model	str	'align'	実行モード（固定）
--img_size	int	256	画像サイズ
--inp_dir	str	''　ダイアログによる指定	変換する画像のあるフォルダ
--out_dir	str	'assets/representative/custom_src_align'	出力するフォルダ
--wing_path	str	'expr/checkpoints/wing.ckpt'
--lm_path	str	'expr/checkpoints/celeba_lm_mean.npz'

コマンド実行例

(py38_learn) python starv2_align.py

StarGAN v2 mode align     Ver 0.01: Starting application...

   --mode           :  align
   --img_size       :  256
   --inp_dir        :  C:/anaconda_win/workspace_2/stargan-v2/assets/representative/custom_src
   --out_dir        :  assets/representative/custom_src_align
   --wing_path      :  expr/checkpoints/wing.ckpt
   --lm_path        :  expr/checkpoints/celeba_lm_mean.npz

Saved the aligned image to f_kiyohara_4.jpg...
Saved the aligned image to f_suzuki_2.jpg...
Saved the aligned image to f_yoshida_1.jpg...
Saved the aligned image to f_yoshinaga_5.jpg...
Saved the aligned image to f_yoshine_1.jpg...
Saved the aligned image to f_yoshioka_5.jpg...
Saved the aligned image to ietake.jpg...
Saved the aligned image to m_kamiki_2.jpg...
Saved the aligned image to m_kusakari_1.jpg...
Saved the aligned image to okegawa.jpg...

processing start >>      2024/08/03 06:12:06
processing end >>        2024/08/03 06:12:08
processing time >>       0:00:01.937256

Finished.

▼「starv2_align.py」ソースコード

# -*- coding: utf-8 -*-
##------------------------------------------
##  StarGAN v2 mode align     Ver 0.01
##
##               2024.07.08 Masahiro Izutsu
##------------------------------------------
## starv2_align.py

import warnings
warnings.simplefilter('ignore')

import argparse
from core.wing import align_faces

import my_imagetool

# Color Escape Code ---------------------------
GREEN = '\033[1;32m'
RED = '\033[1;31m'
NOCOLOR = '\033[0m'
YELLOW = '\033[1;33m'

import os 
import glob

# タイトル
title = 'StarGAN v2 mode align     Ver 0.01'

# 基本情報の表示
def display_info(args):
    print('\n' + GREEN + title + ': Starting application...' + NOCOLOR)
    print('\n   --' + NOCOLOR + 'mode           : ' + NOCOLOR, args.mode)
    print('   --' + YELLOW + 'img_size       : ' + NOCOLOR, args.img_size)
    print('   --' + YELLOW + 'inp_dir        : ' + NOCOLOR, args.inp_dir)
    print('   --' + YELLOW + 'out_dir        : ' + NOCOLOR, args.out_dir)
    print('   --' + YELLOW + 'wing_path      : ' + NOCOLOR, args.wing_path)
    print('   --' + YELLOW + 'lm_path        : ' + NOCOLOR, args.lm_path)
    print(' ')

#-----------------------------------------------

def main(args):

    align_faces(args, args.inp_dir, args.out_dir)


if __name__ == '__main__':
    import my_dialog
    import datetime

    parser = argparse.ArgumentParser()

    # model arguments
    parser.add_argument('--img_size', type=int, default=256,
                        help='Image resolution')

    # misc
    parser.add_argument('--mode', type=str, default='align', help='This argument is used in solver')

    # directory for testing
    parser.add_argument('--inp_dir', type=str, default='',
                        help='input directory when aligning faces')
    parser.add_argument('--out_dir', type=str, default='assets/representative',
                        help='output directory when aligning faces')
    # face alignment
    parser.add_argument('--wing_path', type=str, default='expr/checkpoints/wing.ckpt')
    parser.add_argument('--lm_path', type=str, default='expr/checkpoints/celeba_lm_mean.npz')

    args = parser.parse_args()

    if len(args.inp_dir) == 0:
        filepath = args.inp_dir = my_dialog.select_dir(msg = '元画像　', initdir = './assets/representative')
        if len(filepath) > 1:
            args.inp_dir = filepath
        else:
            exit(0)

    start_time = datetime.datetime.now()        # 時間計測開始

    # 画像ファイルの存在チェック
    file0 = glob.glob(args.inp_dir + '/*.jpg')
    file1 = glob.glob(args.inp_dir + '/*.png')
    if file0 == [] and file1 == []:
        print(RED + f"Image File not found !! '{args.inp_dir}' " + NOCOLOR)
        exit(0)

    # 出力フォルダ名
    base_dir_pair = os.path.split(args.inp_dir)
    path = args.out_dir if args.out_dir[-1] != '/' else args.out_dir[:-1]
    args.out_dir = path + '/' + base_dir_pair[1] + '_align'

    display_info(args)
    main(args)

    # 経過時間
    end_time = datetime.datetime.now()
    print(start_time.strftime('\nprocessing start >>\t %Y/%m/%d %H:%M:%S'))
    print(end_time.strftime('processing end >>\t %Y/%m/%d %H:%M:%S'))
    print('processing time >>\t', end_time - start_time)

    print('\nFinished.\n')

GUI で操作できる StarGAN 実行プログラム「starv2_sample.py」を作成 †

主な機能
・ソース画像と参照画像の入っているフォルダを ダイアログにより選択指定する
　参照画像を横ソース画像を縦に並べた画像が生成される
・3x3以上の枚数の画像がある場合は補間ビデオを同時に生成する

出力ファイルの保存場所とファイル名
・「--result_dir」で指定されたフォルダ内に保存される
・処理結果画像　 → 'reference-' + <ソース画像> + '-' + <参照画像> + '.jpg'
・補間ビデオ　　 → 'video_ref-'+ <ソース画像> + '-' + <参照画像> + '.mp4'

コマンドオプション一覧

コマンドオプション	引数	初期値	意味
--model	str	'align'	実行モード（固定）
--img_size	int	256	画像サイズ
--num_domains	int	2
--latent_dim	int	16
--style_dim	int	64
--w_hpf	int	1
--resume_iter	int	0
--batch_size	int	8
--val_batch_size	int	32
--num_workers	int	4
--seed	int	777
--src_dir	str	''　ダイアログによる指定	ソース画像のあるフォルダ
--ref_dir	str	''　ダイアログによる指定	参照画像のあるフォルダ
--result_dir	str	'expr/results/celeba_hq'	出力するフォルダ
--wing_path	str	'expr/checkpoints/wing.ckpt'

コマンド実行例

(py38_learn) python starv2_sample.py

StarGAN v2 mode sample     Ver 0.01: Starting application...

   --mode           :  sample
   --img_size       :  256
   --num_domains    :  2
   --latent_dim     :  16
   --style_dim      :  64
   --w_hpf          :  1
   --resume_iter    :  0
   --batch_size     :  8
   --val_batch_size :  32
   --num_workers    :  4
   --seed           :  777
   --checkpoint_dir :  expr/checkpoints/celeba_hq
   --src_dir        :  C:/anaconda_win/workspace_2/stargan-v2/assets/representative/celeba_hq/src_3
   --ref_dir        :  C:/anaconda_win/workspace_2/stargan-v2/assets/representative/celeba_hq/ref2_03
   --result_dir     :  expr/results/celeba_hq
   --wing_path      :  expr/checkpoints/wing.ckpt
    :
    :
 Generated file → expr/results/celeba_hq/reference-src_3-ref2_03.jpg
 Generated file → expr/results/celeba_hq/video_ref-src_3-ref2_03.mp4

processing start >>      2024/08/03 06:19:41
processing end >>        2024/08/03 06:20:49
processing time >>       0:01:08.162194

Finished.

・CPU 動作に場合（十数分かかる）

processing start >>      2024/08/03 08:00:57
processing end >>        2024/08/03 08:18:20
processing time >>       0:17:23.289710

▼　- log -　「python starv2_sample.py」

(py38_learn) python starv2_sample.py

StarGAN v2 mode sample     Ver 0.01: Starting application...

   --mode           :  sample
   --img_size       :  256
   --num_domains    :  2
   --latent_dim     :  16
   --style_dim      :  64
   --w_hpf          :  1
   --resume_iter    :  0
   --batch_size     :  8
   --val_batch_size :  32
   --num_workers    :  4
   --seed           :  777
   --checkpoint_dir :  expr/checkpoints/celeba_hq
   --src_dir        :  C:/anaconda_win/workspace_2/stargan-v2/assets/representative/celeba_hq/src_3
   --ref_dir        :  C:/anaconda_win/workspace_2/stargan-v2/assets/representative/celeba_hq/ref2_03
   --result_dir     :  expr/results/celeba_hq
   --wing_path      :  expr/checkpoints/wing.ckpt

Number of parameters of generator: 43467395
Number of parameters of mapping_network: 2438272
Number of parameters of style_encoder: 20916928
Number of parameters of discriminator: 20852290
Number of parameters of fan: 6333603
Initializing generator...
Initializing mapping_network...
Initializing style_encoder...
Initializing discriminator...
Preparing DataLoader for the generation phase...
Preparing DataLoader for the generation phase...
Loading checkpoint from expr/checkpoints/celeba_hq\100000_nets_ema.ckpt...
Working on expr/results/celeba_hq\reference.jpg...
Working on expr/results/celeba_hq\video_ref.mp4...
video_ref: 100%|███████████████████████████████████████████████████████████████████████| 10/10 [00:18<00:00,  1.87s/it]
writing video to expr/results/celeba_hq\video_ref.mp4:   0%|                                   | 0/258 [00:00<?, ?it/s]ffmpeg version 7.0.1 Copyright (c) 2000-2024 the FFmpeg developers
  built with clang version 18.1.8
  configuration: --prefix=/d/bld/ffmpeg_1719926247333/_h_env/Library --cc=clang.exe --cxx=clang++.exe --nm=llvm-nm --ar=llvm-ar --disable-doc --disable-openssl --enable-demuxer=dash --enable-hardcoded-tables --enable-libfreetype --enable-libharfbuzz --enable-libfontconfig --enable-libopenh264 --enable-libdav1d --ld=lld-link --target-os=win64 --enable-cross-compile --toolchain=msvc --host-cc=clang.exe --extra-libs=ucrt.lib --extra-libs=vcruntime.lib --extra-libs=oldnames.lib --strip=llvm-strip --disable-stripping --host-extralibs= --disable-libopenvino --enable-gpl --enable-libx264 --enable-libx265 --enable-libaom --enable-libsvtav1 --enable-libxml2 --enable-pic --enable-shared --disable-static --enable-version3 --enable-zlib --enable-libopus --pkg-config=/d/bld/ffmpeg_1719926247333/_build_env/Library/bin/pkg-config
  libavutil      59.  8.100 / 59.  8.100
  libavcodec     61.  3.100 / 61.  3.100
  libavformat    61.  1.100 / 61.  1.100
  libavdevice    61.  1.100 / 61.  1.100
  libavfilter    10.  1.100 / 10.  1.100
  libswscale      8.  1.100 /  8.  1.100
  libswresample   5.  1.100 /  5.  1.100
  libpostproc    58.  1.100 / 58.  1.100
Input #0, rawvideo, from 'pipe:':
  Duration: N/A, start: 0.000000, bitrate: 874905 kb/s
  Stream #0:0: Video: rawvideo (RGB[24] / 0x18424752), rgb24, 2848x512, 874905 kb/s, 25 tbr, 25 tbn
Stream mapping:
  Stream #0:0 (rawvideo) -> setpts:default
  setpts:default -> Stream #0:0 (libx264)
[libx264 @ 000002236F5A82C0] using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
[libx264 @ 000002236F5A82C0] profile High, level 4.0, 4:2:0, 8-bit
[libx264 @ 000002236F5A82C0] 264 - core 164 r3095 baee400 - H.264/MPEG-4 AVC codec - Copyleft 2003-2022 - http://www.videolan.org/x264.html - options: cabac=1 ref=3 deblock=1:0:0 analyse=0x3:0x113 me=hex subme=7 psy=1 psy_rd=1.00:0.00 mixed_ref=1 me_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 fast_pskip=1 chroma_qp_offset=-2 threads=16 lookahead_threads=2 sliced_threads=0 nr=0 decimate=1 interlaced=0 bluray_compat=0 constrained_intra=0 bframes=3 b_pyramid=2 b_adapt=1 b_bias=0 direct=1 weightb=1 open_gop=0 weightp=2 keyint=250 keyint_min=25 scenecut=40 intra_refresh=0 rc_lookahead=40 rc=crf mbtree=1 crf=23.0 qcomp=0.60 qpmin=0 qpmax=69 qpstep=4 ip_ratio=1.40 aq=1:1.00
Output #0, mp4, to 'expr/results/celeba_hq\video_ref.mp4':
  Metadata:
    encoder         : Lavf61.1.100
  Stream #0:0: Video: h264 (avc1 / 0x31637661), yuv420p(progressive), 2848x512, q=2-31, 30 fps, 15360 tbn
      Metadata:
        encoder         : Lavc61.3.100 libx264
      Side data:
        cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: N/A
writing video to expr/results/celeba_hq\video_ref.mp4: 100%|█████████████████████████| 258/258 [00:03<00:00, 78.14it/s]
[out#0/mp4 @ 000002236F58E480] video:4012KiB audio:0KiB subtitle:0KiB other streams:0KiB global headers:0KiB muxing overhead: 0.195275%
frame=  619 fps=183 q=-1.0 Lsize=    4020KiB time=00:00:20.56 bitrate=1601.3kbits/s dup=361 drop=0 speed= 6.1x
[libx264 @ 000002236F5A82C0] frame I:3     Avg QP:18.07  size:173891
[libx264 @ 000002236F5A82C0] frame P:185   Avg QP:19.25  size: 17832
[libx264 @ 000002236F5A82C0] frame B:431   Avg QP:20.91  size:   667
[libx264 @ 000002236F5A82C0] consecutive B-frames:  0.5% 13.6% 19.4% 66.6%
[libx264 @ 000002236F5A82C0] mb I  I16..4:  7.2% 74.0% 18.8%
[libx264 @ 000002236F5A82C0] mb P  I16..4:  1.3%  4.0%  1.1%  P16..4: 14.5%  5.7%  3.7%  0.0%  0.0%    skip:69.7%
[libx264 @ 000002236F5A82C0] mb B  I16..4:  0.0%  0.1%  0.0%  B16..8:  5.6%  0.1%  0.1%  direct: 0.3%  skip:93.8%  L0:33.7% L1:60.6% BI: 5.7%
[libx264 @ 000002236F5A82C0] 8x8 transform intra:65.1% inter:64.8%
[libx264 @ 000002236F5A82C0] coded y,uvDC,uvAC intra: 63.0% 69.7% 38.7% inter: 4.3% 4.5% 0.4%
[libx264 @ 000002236F5A82C0] i16 v,h,dc,p: 58% 19%  8% 15%
[libx264 @ 000002236F5A82C0] i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 25% 11% 19%  6%  7% 10%  5% 10%  8%
[libx264 @ 000002236F5A82C0] i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 26% 11% 13%  6% 11% 15%  6%  9%  4%
[libx264 @ 000002236F5A82C0] i8c dc,h,v,p: 41% 19% 33%  7%
[libx264 @ 000002236F5A82C0] Weighted P-Frames: Y:0.0% UV:0.0%
[libx264 @ 000002236F5A82C0] ref P L0: 63.7% 30.1%  5.4%  0.8%
[libx264 @ 000002236F5A82C0] ref B L0: 96.5%  3.2%  0.4%
[libx264 @ 000002236F5A82C0] ref B L1: 97.8%  2.2%
[libx264 @ 000002236F5A82C0] kb/s:1592.71
 Generated file → expr/results/celeba_hq/reference-src_3-ref2_03.jpg
 Generated file → expr/results/celeba_hq/video_ref-src_3-ref2_03.mp4

processing start >>      2024/08/03 06:19:41
processing end >>        2024/08/03 06:20:49
processing time >>       0:01:08.162194

Finished.

※「ffmpeg」ログ出力を抑制すると途中で止まってしまうのでそのままにしている

生成された補間ビデオ
　

▼「starv2_sample.py」ソースコード

# -*- coding: utf-8 -*-
##------------------------------------------
##  StarGAN v2 mode sample     Ver 0.01
##
##               2024.07.08 Masahiro Izutsu
##------------------------------------------
## starv2_sample.py

import warnings
warnings.simplefilter('ignore')

import os
import argparse
from munch import Munch
from torch.backends import cudnn
import torch

from core.data_loader import get_test_loader
from core.solver import Solver
import shutil
import my_imagetool

# Color Escape Code ---------------------------
GREEN = '\033[1;32m'
RED = '\033[1;31m'
NOCOLOR = '\033[0m'
YELLOW = '\033[1;33m'
CYAN = '\033[1;36m'
BLUE = '\033[1;34m'

# タイトル
title = 'StarGAN v2 mode sample     Ver 0.01'

# 基本情報の表示
def display_info(args):
    print('\n' + GREEN + title + ': Starting application...' + NOCOLOR)
    print('\n   --' + NOCOLOR + 'mode           : ' + NOCOLOR, args.mode)
    print('   --' + YELLOW + 'img_size       : ' + NOCOLOR, args.img_size)
    print('   --' + YELLOW + 'num_domains    : ' + NOCOLOR, args.num_domains)
    print('   --' + YELLOW + 'latent_dim     : ' + NOCOLOR, args.latent_dim)
    print('   --' + YELLOW + 'style_dim      : ' + NOCOLOR, args.style_dim)
    print('   --' + YELLOW + 'w_hpf          : ' + NOCOLOR, args.w_hpf)
    print('   --' + YELLOW + 'resume_iter    : ' + NOCOLOR, args.resume_iter)
    print('   --' + YELLOW + 'batch_size     : ' + NOCOLOR, args.batch_size)
    print('   --' + YELLOW + 'val_batch_size : ' + NOCOLOR, args.val_batch_size)
    print('   --' + YELLOW + 'num_workers    : ' + NOCOLOR, args.num_workers)
    print('   --' + YELLOW + 'seed           : ' + NOCOLOR, args.seed)
    print('   --' + CYAN + 'checkpoint_dir : ' + NOCOLOR, args.checkpoint_dir)
    print('   --' + CYAN + 'src_dir        : ' + NOCOLOR, args.src_dir)
    print('   --' + CYAN + 'ref_dir        : ' + NOCOLOR, args.ref_dir)
    print('   --' + CYAN + 'result_dir     : ' + NOCOLOR, args.result_dir)
    print('   --' + YELLOW + 'wing_path      : ' + NOCOLOR, args.wing_path)
    print(' ')

#-----------------------------------------------

def subdirs(dname):
    return [d for d in os.listdir(dname)
            if os.path.isdir(os.path.join(dname, d))]

def get_destination_path(spath, args):
    root_ext_pair = os.path.splitext(spath)
    dir0 = os.path.basename(args.src_dir)
    dir1 = os.path.basename(args.ref_dir)

    dpath = root_ext_pair[0] + '-' + dir0 + '-' + dir1 + root_ext_pair[1]
    return dpath


def main(args):
    cudnn.benchmark = True
    torch.manual_seed(args.seed)

    solver = Solver(args)

    assert len(subdirs(args.src_dir)) == args.num_domains
    assert len(subdirs(args.ref_dir)) == args.num_domains
    loaders = Munch(src=get_test_loader(root=args.src_dir, img_size=args.img_size, batch_size=args.val_batch_size, shuffle=False, num_workers=args.num_workers),
                ref=get_test_loader(root=args.ref_dir, img_size=args.img_size, batch_size=args.val_batch_size, shuffle=False, num_workers=args.num_workers))
    solver.sample(loaders)

    # 'reference.jpg' を別名でコピーする
    spath = args.result_dir + '/reference.jpg'
    dpath = get_destination_path(spath, args)
    shutil.copy(spath, dpath)
    print(f' Generated file → {dpath}')
    my_imagetool.image2disp(filepath = dpath, maxsize = my_imagetool.WINDOW_WIDTH)

    spath = args.result_dir + '/video_ref.mp4'
    if os.path.isfile(spath):
        dpath = get_destination_path(spath, args)
        shutil.copy(spath, dpath)
        print(f' Generated file → {dpath}')
        my_imagetool.image2disp(filepath = dpath, maxsize = my_imagetool.WINDOW_WIDTH)

if __name__ == '__main__':
    import my_dialog
    import datetime

    parser = argparse.ArgumentParser()

    # model arguments
    parser.add_argument('--img_size', type=int, default=256, help='Image resolution')
    parser.add_argument('--num_domains', type=int, default=2, help='Number of domains')
    parser.add_argument('--latent_dim', type=int, default=16, help='Latent vector dimension')
    parser.add_argument('--style_dim', type=int, default=64, help='Style code dimension')

    # weight for objective functions
    parser.add_argument('--w_hpf', type=float, default=1, help='weight for high-pass filtering')

    # training arguments
    parser.add_argument('--resume_iter', type=int, default=0, help='Iterations to resume training/testing')
    parser.add_argument('--batch_size', type=int, default=8, help='Batch size for training')
    parser.add_argument('--val_batch_size', type=int, default=32, help='Batch size for validation')

    # misc
    parser.add_argument('--mode', type=str, default='sample', help='This argument is used in solver')
    parser.add_argument('--num_workers', type=int, default=4, help='Number of workers used in DataLoader')
    parser.add_argument('--seed', type=int, default=777, help='Seed for random number generator')

    # directory for training
    parser.add_argument('--checkpoint_dir', type=str, default='expr/checkpoints/celeba_hq', help='Directory for saving network checkpoints')

    # directory for testing
    parser.add_argument('--src_dir', type=str, default='', help='Directory containing input source images')
    parser.add_argument('--ref_dir', type=str, default='', help='Directory containing input reference images')
    parser.add_argument('--result_dir', type=str, default='expr/results/celeba_hq', help='Directory for saving generated images and videos')

    # face alignment
    parser.add_argument('--wing_path', type=str, default='expr/checkpoints/wing.ckpt')

    args = parser.parse_args()

    if len(args.src_dir) == 0:
        filepath = args.src_dir = my_dialog.select_dir(msg = '元画像(source)　', initdir = './assets/representative/celeba_hq')
        if len(filepath) > 1:
            args.src_dir = filepath
        else:
            exit(0)

    if len(args.ref_dir) == 0:
        filepath = args.ref_dir = my_dialog.select_dir(msg = '参照画像(reference)　', initdir = './assets/representative/celeba_hq')
        if len(filepath) > 1:
            args.ref_dir = filepath
        else:
            exit(0)

    display_info(args)

    start_time = datetime.datetime.now()        # 時間計測開始
    main(args)

    # 経過時間
    end_time = datetime.datetime.now()
    print(start_time.strftime('\nprocessing start >>\t %Y/%m/%d %H:%M:%S'))
    print(end_time.strftime('processing end >>\t %Y/%m/%d %H:%M:%S'))
    print('processing time >>\t', end_time - start_time)

    print('\nFinished.\n')

対処した問題点とエラー詳細 †

FileNotFoundError: [WinError 3] 指定されたパスが見つかりません。: '{:' †

エラー内容

▼　FileNotFoundError: ログ

(py38_gan) python main.py --mode sample --num_domains 2 --resume_iter 100000 --w_hpf 1 --checkpoint_dir expr/checkpoints/celeba_hq --result_dir expr/results/celeba_hq --src_dir assets/representative/celeba_hq/src_0 --ref_dir assets/representative/celeba_hq/ref_0
Namespace(batch_size=8, beta1=0.0, beta2=0.99, checkpoint_dir='expr/checkpoints/celeba_hq', ds_iter=100000, eval_dir='expr/eval', eval_every=50000, f_lr=1e-06, hidden_dim=512, img_size=256, inp_dir='assets/representative/custom/female', lambda_cyc=1, lambda_ds=1, lambda_reg=1, lambda_sty=1, latent_dim=16, lm_path='expr/checkpoints/celeba_lm_mean.npz', lr=0.0001, mode='sample', num_domains=2, num_outs_per_domain=10, num_workers=4, out_dir='assets/representative/celeba_hq/src/female', print_every=10, randcrop_prob=0.5, ref_dir='assets/representative/celeba_hq/ref_0', result_dir='expr/results/celeba_hq', resume_iter=100000, sample_dir='expr/samples', sample_every=5000, save_every=10000, seed=777, src_dir='assets/representative/celeba_hq/src_0', style_dim=64, total_iters=100000, train_img_dir='data/celeba_hq/train', val_batch_size=32, val_img_dir='data/celeba_hq/val', w_hpf=1.0, weight_decay=0.0001, wing_path='expr/checkpoints/wing.ckpt')
Number of parameters of generator: 43467395
Number of parameters of mapping_network: 2438272
Number of parameters of style_encoder: 20916928
Number of parameters of discriminator: 20852290
Number of parameters of fan: 6333603
Traceback (most recent call last):
  File "main.py", line 182, in <module>
    main(args)
  File "main.py", line 37, in main
    solver = Solver(args)
  File "N:\anaconda_win\work\stargan-v2\core\solver.py", line 58, in __init__
    self.ckptios = [CheckpointIO(ospj(args.checkpoint_dir, '{:06d}_nets_ema.ckpt'), data_parallel=True, **self.nets_ema)]
  File "N:\anaconda_win\work\stargan-v2\core\checkpoint.py", line 17, in __init__
    os.makedirs(os.path.dirname(fname_template), exist_ok=True)
  File "C:\Users\izuts\anaconda3\envs\py38_gan\lib\os.py", line 223, in makedirs
    mkdir(name, mode)
FileNotFoundError: [WinError 3] 指定されたパスが見つかりません。: '{:'

参考サイト
Path error upon running the example you provided to transform a custom image #3

対処法：「stargan-v2/core/solver.py」→ 58行目変更

        else:
#            self.ckptios = [CheckpointIO(ospj(args.checkpoint_dir, '{:06d}_nets_ema.ckpt'), data_parallel=True, **self.nets_ema)]
            self.ckptios = [CheckpointIO(ospj(args.checkpoint_dir, '100000_nets_ema.ckpt'), data_parallel=True, **self.nets_ema)]

RuntimeError: Error(s) in loading state_dict for Generator: Missing key(s) in state_dict: "hpf.filter" †

エラー内容

▼　RuntimeError: ログ

(py38_gan) python main.py --mode sample --num_domains 2 --resume_iter 100000 --w_hpf 1 --checkpoint_dir expr/checkpoints/celeba_hq --result_dir expr/results/celeba_hq --src_dir assets/representative/celeba_hq/src_0 --ref_dir assets/representative/celeba_hq/ref_0
Namespace(batch_size=8, beta1=0.0, beta2=0.99, checkpoint_dir='expr/checkpoints/celeba_hq', ds_iter=100000, eval_dir='expr/eval', eval_every=50000, f_lr=1e-06, hidden_dim=512, img_size=256, inp_dir='assets/representative/custom/female', lambda_cyc=1, lambda_ds=1, lambda_reg=1, lambda_sty=1, latent_dim=16, lm_path='expr/checkpoints/celeba_lm_mean.npz', lr=0.0001, mode='sample', num_domains=2, num_outs_per_domain=10, num_workers=4, out_dir='assets/representative/celeba_hq/src/female', print_every=10, randcrop_prob=0.5, ref_dir='assets/representative/celeba_hq/ref_0', result_dir='expr/results/celeba_hq', resume_iter=100000, sample_dir='expr/samples', sample_every=5000, save_every=10000, seed=777, src_dir='assets/representative/celeba_hq/src_0', style_dim=64, total_iters=100000, train_img_dir='data/celeba_hq/train', val_batch_size=32, val_img_dir='data/celeba_hq/val', w_hpf=1.0, weight_decay=0.0001, wing_path='expr/checkpoints/wing.ckpt')
Number of parameters of generator: 43467395
Number of parameters of mapping_network: 2438272
Number of parameters of style_encoder: 20916928
Number of parameters of discriminator: 20852290
Number of parameters of fan: 6333603
Initializing generator...
Initializing mapping_network...
Initializing style_encoder...
Initializing discriminator...
Preparing DataLoader for the generation phase...
Preparing DataLoader for the generation phase...
Loading checkpoint from expr/checkpoints/celeba_hq\100000_nets_ema.ckpt...
Traceback (most recent call last):
  File "main.py", line 182, in <module>
    main(args)
  File "main.py", line 73, in main
    solver.sample(loaders)
  File "C:\Users\izuts\anaconda3\envs\py38_gan\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)
  File "N:\anaconda_win\work\stargan-v2\core\solver.py", line 179, in sample
    self._load_checkpoint(args.resume_iter)
  File "N:\anaconda_win\work\stargan-v2\core\solver.py", line 74, in _load_checkpoint
    ckptio.load(step)
  File "N:\anaconda_win\work\stargan-v2\core\checkpoint.py", line 48, in load
    module.module.load_state_dict(module_dict[name])
  File "C:\Users\izuts\anaconda3\envs\py38_gan\lib\site-packages\torch\nn\modules\module.py", line 2152, in load_state_dict
    raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for Generator:
        Missing key(s) in state_dict: "hpf.filter".

参考サイト
Error when loading state_dict for Generator #103

対処法：「stargan-v2/core/checkpoint.py」→ 48行目変更

            if self.data_parallel:
#                module.module.load_state_dict(module_dict[name])
                module.module.load_state_dict(module_dict[name], False)

UnboundLocalError: local variable 'frames' referenced before assignment †

エラー内容

▼　UnboundLocalError: ログ

(py38_gan) python main.py --mode sample --num_domains 2 --resume_iter 100000 --w_hpf 1 --checkpoint_dir expr/checkpoints/celeba_hq --result_dir expr/results/celeba_hq --src_dir assets/representative/celeba_hq/src_0 --ref_dir assets/representative/celeba_hq/ref_0
Namespace(batch_size=8, beta1=0.0, beta2=0.99, checkpoint_dir='expr/checkpoints/celeba_hq', ds_iter=100000, eval_dir='expr/eval', eval_every=50000, f_lr=1e-06, hidden_dim=512, img_size=256, inp_dir='assets/representative/custom/female', lambda_cyc=1, lambda_ds=1, lambda_reg=1, lambda_sty=1, latent_dim=16, lm_path='expr/checkpoints/celeba_lm_mean.npz', lr=0.0001, mode='sample', num_domains=2, num_outs_per_domain=10, num_workers=4, out_dir='assets/representative/celeba_hq/src/female', print_every=10, randcrop_prob=0.5, ref_dir='assets/representative/celeba_hq/ref_0', result_dir='expr/results/celeba_hq', resume_iter=100000, sample_dir='expr/samples', sample_every=5000, save_every=10000, seed=777, src_dir='assets/representative/celeba_hq/src_0', style_dim=64, total_iters=100000, train_img_dir='data/celeba_hq/train', val_batch_size=32, val_img_dir='data/celeba_hq/val', w_hpf=1.0, weight_decay=0.0001, wing_path='expr/checkpoints/wing.ckpt')
Number of parameters of generator: 43467395
Number of parameters of mapping_network: 2438272
Number of parameters of style_encoder: 20916928
Number of parameters of discriminator: 20852290
Number of parameters of fan: 6333603
Initializing generator...
Initializing mapping_network...
Initializing style_encoder...
Initializing discriminator...
Preparing DataLoader for the generation phase...
Preparing DataLoader for the generation phase...
Loading checkpoint from expr/checkpoints/celeba_hq\100000_nets_ema.ckpt...
Working on expr/results/celeba_hq\reference.jpg...
Working on expr/results/celeba_hq\video_ref.mp4...
video_ref: 100%|█████████████████████████████████████████| 2/2 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "main.py", line 182, in <module>
    main(args)
  File "main.py", line 73, in main
    solver.sample(loaders)
  File "C:\Users\izuts\anaconda3\envs\py38_gan\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)
  File "N:\anaconda_win\work\stargan-v2\core\solver.py", line 190, in sample
    utils.video_ref(nets_ema, args, src.x, ref.x, ref.y, fname)
  File "C:\Users\izuts\anaconda3\envs\py38_gan\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)
  File "N:\anaconda_win\work\stargan-v2\core\utils.py", line 222, in video_ref
    video.append(frames[-1:])
UnboundLocalError: local variable 'frames' referenced before assignment

対処法1：「stargan-v2/core/utils.py」→ 202行目追加

def video_ref(nets, args, x_src, x_ref, y_ref, fname):
    video = []
    frames = []

対処法2：「stargan-v2/core/solver.py」→ 188行目から変更
生成画像を保存した後、保管ビデオの作成時にエラーとなる（作成には 3つ以上の画像が必要）

        if len(ref.x) > 2:
            fname = ospj(args.result_dir, 'video_ref.mp4')
            print('Working on {}...'.format(fname))
            utils.video_ref(nets_ema, args, src.x, ref.x, ref.y, fname)

TypeError: gaussian() got an unexpected keyword argument 'multichannel' †

エラー内容

▼　TypeError: ログ

(py38_gan) python main.py --mode align --inp_dir assets/representative/custom_1/male --out_dir assets/representative/celeba_hq/ref_test/male
Namespace(batch_size=8, beta1=0.0, beta2=0.99, checkpoint_dir='expr/checkpoints', ds_iter=100000, eval_dir='expr/eval', eval_every=50000, f_lr=1e-06, hidden_dim=512, img_size=256, inp_dir='assets/representative/custom_1/male', lambda_cyc=1, lambda_ds=1, lambda_reg=1, lambda_sty=1, latent_dim=16, lm_path='expr/checkpoints/celeba_lm_mean.npz', lr=0.0001, mode='align', num_domains=2, num_outs_per_domain=10, num_workers=4, out_dir='assets/representative/celeba_hq/ref_test/male', print_every=10, randcrop_prob=0.5, ref_dir='assets/representative/celeba_hq/ref', result_dir='expr/results', resume_iter=0, sample_dir='expr/samples', sample_every=5000, save_every=10000, seed=777, src_dir='assets/representative/celeba_hq/src', style_dim=64, total_iters=100000, train_img_dir='data/celeba_hq/train', val_batch_size=32, val_img_dir='data/celeba_hq/val', w_hpf=1, weight_decay=0.0001, wing_path='expr/checkpoints/wing.ckpt')
Number of parameters of generator: 43467395
Number of parameters of mapping_network: 2438272
Number of parameters of style_encoder: 20916928
Number of parameters of discriminator: 20852290
Number of parameters of fan: 6333603
Initializing generator...
Initializing mapping_network...
Initializing style_encoder...
Initializing discriminator...
Traceback (most recent call last):
  File "main.py", line 182, in <module>
    main(args)
  File "main.py", line 78, in main
    align_faces(args, args.inp_dir, args.out_dir)
  File "N:\anaconda_win\work\stargan-v2\core\wing.py", line 427, in align_faces
    x_aligned = aligner.align(x)
  File "N:\anaconda_win\work\stargan-v2\core\wing.py", line 305, in align
    img_np, landmarks = pad_mirror(img_np, landmarks)
  File "N:\anaconda_win\work\stargan-v2\core\wing.py", line 392, in pad_mirror
    small_blurred = gaussian(cv2.resize(img, (W, H)), H//100, multichannel=True)
TypeError: gaussian() got an unexpected keyword argument 'multichannel'

参考サイト
TypeError: rescale() got an unexpected keyword argument 'multichannel' #6
scikit-image API reference

対処法：「stargan-v2/core/wing.py」→ 392行目変更

#    small_blurred = gaussian(cv2.resize(img, (W, H)), H//100, multichannel=True)
    small_blurred = gaussian(cv2.resize(img, (W, H)), H//100, channel_axis=1)

更新履歴 †

2024/07/03 初版

参考資料 †

StarGAN

Python
- [Python] assert文で超お手軽テスト
- Pythonで例外を発生させる：raise

Error

Others
- 画像変換の最新動向【2018-2020年】