Submit Search
RUTILEA社内勉強会第6回 「超解像」
Download as PPTX, PDF
2 likes
264 views
T
TRUE_RUTILEA
RUTILEA社内で定期的に行っている勉強会の第6回です. 第6回のテーマは”超解像”です.
Engineering
Read more
1 of 22
Download now
Download to read offline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
More Related Content
PDF
兼本研究室プログラム講義4回目
ytanno
PPTX
RUTILEA社内勉強会第8回 「シリアル通信」
TRUE_RUTILEA
PPTX
RUTILEA社内勉強会第7回 「LSTMを用いた異常検知」
TRUE_RUTILEA
PDF
RUTILEA社内勉強会第5回 「強化学習」
TRUE_RUTILEA
PDF
RUTILEA社内勉強会第4回 「敵対的生成ネットワーク(GAN)」
TRUE_RUTILEA
PPTX
RUTILEA社内勉強会第2回 「多重検定」
TRUE_RUTILEA
PDF
RUTILEA社内勉強会第1回 「転移学習」
TRUE_RUTILEA
PDF
RUTILEA社内勉強会第3回 「意外な統計」
TRUE_RUTILEA
兼本研究室プログラム講義4回目
ytanno
RUTILEA社内勉強会第8回 「シリアル通信」
TRUE_RUTILEA
RUTILEA社内勉強会第7回 「LSTMを用いた異常検知」
TRUE_RUTILEA
RUTILEA社内勉強会第5回 「強化学習」
TRUE_RUTILEA
RUTILEA社内勉強会第4回 「敵対的生成ネットワーク(GAN)」
TRUE_RUTILEA
RUTILEA社内勉強会第2回 「多重検定」
TRUE_RUTILEA
RUTILEA社内勉強会第1回 「転移学習」
TRUE_RUTILEA
RUTILEA社内勉強会第3回 「意外な統計」
TRUE_RUTILEA
Featured
(20)
PDF
2024 Trend Updates: What Really Works In SEO & Content Marketing
Search Engine Journal
PDF
Storytelling For The Web: Integrate Storytelling in your Design Process
Chiara Aliotta
PDF
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
OECD Directorate for Financial and Enterprise Affairs
PDF
How to Leverage AI to Boost Employee Wellness - Lydia Di Francesco - SocialHR...
SocialHRCamp
PDF
2024 State of Marketing Report – by Hubspot
Marius Sescu
PDF
Everything You Need To Know About ChatGPT
Expeed Software
PDF
Product Design Trends in 2024 | Teenage Engineerings
Pixeldarts
PDF
How Race, Age and Gender Shape Attitudes Towards Mental Health
ThinkNow
PDF
AI Trends in Creative Operations 2024 by Artwork Flow.pdf
marketingartwork
PDF
Skeleton Culture Code
Skeleton Technologies
PDF
PEPSICO Presentation to CAGNY Conference Feb 2024
Neil Kimberley
PDF
Content Methodology: A Best Practices Report (Webinar)
contently
PPTX
How to Prepare For a Successful Job Search for 2024
Albert Qian
PDF
Social Media Marketing Trends 2024 // The Global Indie Insights
Kurio // The Social Media Age(ncy)
PDF
Trends In Paid Search: Navigating The Digital Landscape In 2024
Search Engine Journal
PDF
5 Public speaking tips from TED - Visualized summary
SpeakerHub
PDF
ChatGPT and the Future of Work - Clark Boyd
Clark Boyd
PDF
Getting into the tech field. what next
Tessa Mero
PDF
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Lily Ray
PDF
How to have difficult conversations
Rajiv Jayarajah, MAppComm, ACC
2024 Trend Updates: What Really Works In SEO & Content Marketing
Search Engine Journal
Storytelling For The Web: Integrate Storytelling in your Design Process
Chiara Aliotta
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
OECD Directorate for Financial and Enterprise Affairs
How to Leverage AI to Boost Employee Wellness - Lydia Di Francesco - SocialHR...
SocialHRCamp
2024 State of Marketing Report – by Hubspot
Marius Sescu
Everything You Need To Know About ChatGPT
Expeed Software
Product Design Trends in 2024 | Teenage Engineerings
Pixeldarts
How Race, Age and Gender Shape Attitudes Towards Mental Health
ThinkNow
AI Trends in Creative Operations 2024 by Artwork Flow.pdf
marketingartwork
Skeleton Culture Code
Skeleton Technologies
PEPSICO Presentation to CAGNY Conference Feb 2024
Neil Kimberley
Content Methodology: A Best Practices Report (Webinar)
contently
How to Prepare For a Successful Job Search for 2024
Albert Qian
Social Media Marketing Trends 2024 // The Global Indie Insights
Kurio // The Social Media Age(ncy)
Trends In Paid Search: Navigating The Digital Landscape In 2024
Search Engine Journal
5 Public speaking tips from TED - Visualized summary
SpeakerHub
ChatGPT and the Future of Work - Clark Boyd
Clark Boyd
Getting into the tech field. what next
Tessa Mero
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Lily Ray
How to have difficult conversations
Rajiv Jayarajah, MAppComm, ACC
Ad
RUTILEA社内勉強会第6回 「超解像」
1.
超解像 2020/02/06 田中 達也
2.
超解像とは 画像の解像度を上げる画像処理技術 参照 http://www.togawa.cs.waseda.ac.jp/research/system/system.html
3.
超解像 画像を拡大させた時、間のデータを既知のデータから補間する https://qiita.com/jiny2001/items/e2175b52013bf655d617
4.
学習を用いない超解像技術 Nearest neighbor 最も近くの点を用いる Linear 近傍2点から線形に補間 する Cubic 近傍4点から、三次式で 補間する Bicubic interpolation
- Wikipedia
5.
試した手法 • Bilinear • FCBI •
RAISR • SRCNN 学習を用いない
6.
線形補間(Bilinear) 横方向に2画素の平均値で埋める 縦方向に2画素の平均値で埋める 1 3 5 3 1
2 3 5 4 3 1 2 3 3 3 3 5 4 3
7.
線形補間 5回繰り返し行った 画像がボケてしまう
8.
FCBI エッジをなるべく残すように4つのうち2つを選択して、平均をとる。 手順 ① ピクセルを拡げる ② 斜め方向からピクセルを埋める ③
縦横方向からピクセルを埋める 𝒍 𝟏 ③ 𝒍 𝟐 ③ ② ③ 𝒍 𝟑 ③ 𝒍 𝟒
9.
FCBI 平均をとる2つ画素の選び方 エッジ判定を行う |𝒍 𝟏-𝒍 𝟒|<TM
かつ|𝒍 𝟐-𝒍 𝟑|<TM かつ | (𝒍 𝟏+𝒍 𝟒)/2 - (𝒍 𝟐+𝒍 𝟑)/2 | < TM ならば 非エッジと判定 i) エッジの場合 その差の少ない2ピクセルを線形補間 ii) 非エッジの場合 補間したいピクセルの周辺の画素をh1とh2の重み付けで 計算を行い、h1 と h2 の大小で勾配の変化の小さい向き を判定. h1 h2 https://blog.awm.jp/2017/03/08/fcbi/ 𝒍 𝟏 ③ 𝒍 𝟐 ③ ② ③ 𝒍 𝟑 ③ 𝒍 𝟒
10.
FCBI
11.
機械学習を用いた超解像 • RAISR Googleが出した論文 処理速度が速い ディープラーニングの手法を使っていない
12.
RAISR 学習時 RAISR: Rapid and
Accurate Image Super Resolution
13.
RAISR: Rapid and
Accurate Image Super Resolution 実行時 RAISR
14.
元画像 RAISR RAISR
15.
比較画像
16.
RAISR 補正前 補正後
17.
SRCNN ディープラーニングを超解像に応用した最初の例 𝑓1 = 5 𝑓2
= 1 𝑓3 = 9 𝑛1 = 64 𝑛2 = 32
18.
• データセット Google Drive
からダウンロード train : 90枚 val : 10枚 test : 10枚 入力画像Y(低解像度)・・・元画像を縮小し、Bicubic補間で戻す 教師画像X(高解像度)・・・元画像
19.
パッチを抽出し、各バッチを高次元ベクトル(64次元)として表す。 各高次元ベクトル(64次元)を別の高次元ベクトル(32次元)にマッピング パッチごとの表現を集約し、高解像度画像を生成 1層目 2層目 3層目 𝑊1: 𝑐 ×
𝑓1 × 𝑓1 × 𝑛1 𝑊2: 𝑐 × 𝑓2 × 𝑓2 × 𝑛2 𝐵1: 𝑛1 𝐵2: 𝑛2 SRCNNの各層の役割 𝑓1 = 5 𝑓2 = 1 𝑓3 = 9 𝑛1 = 64 𝑛2 = 32
20.
損失関数 最小二乗誤差(MSE) X : 高解像度画像 Y
: 低解像度画像
21.
補正前 補正後 SRCNN
22.
まとめ • RAISRは高速だが、ダイカストの画像に適応するには 性能が不十分かもしれない • CNNによる超解像が一番うまくいきそう •
今後、SRCNNの派生形(層を増やす、GANの適応)を試 していきたい
Download