陳昕璇
    • Create new note
    • Create a note from template
      • Sharing URL Link copied
      • /edit
      • View mode
        • Edit mode
        • View mode
        • Book mode
        • Slide mode
        Edit mode View mode Book mode Slide mode
      • Customize slides
      • Note Permission
      • Read
        • Only me
        • Signed-in users
        • Everyone
        Only me Signed-in users Everyone
      • Write
        • Only me
        • Signed-in users
        • Everyone
        Only me Signed-in users Everyone
      • Engagement control Commenting, Suggest edit, Emoji Reply
    • Invite by email
      Invitee

      This note has no invitees

    • Publish Note

      Share your work with the world Congratulations! 🎉 Your note is out in the world Publish Note No publishing access yet

      Your note will be visible on your profile and discoverable by anyone.
      Your note is now live.
      This note is visible on your profile and discoverable online.
      Everyone on the web can find and read all notes of this public team.

      Your account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

      Your team account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

      Explore these features while you wait
      Complete general settings
      Bookmark and like published notes
      Write a few more notes
      Complete general settings
      Write a few more notes
      See published notes
      Unpublish note
      Please check the box to agree to the Community Guidelines.
      View profile
    • Commenting
      Permission
      Disabled Forbidden Owners Signed-in users Everyone
    • Enable
    • Permission
      • Forbidden
      • Owners
      • Signed-in users
      • Everyone
    • Suggest edit
      Permission
      Disabled Forbidden Owners Signed-in users Everyone
    • Enable
    • Permission
      • Forbidden
      • Owners
      • Signed-in users
    • Emoji Reply
    • Enable
    • Versions and GitHub Sync
    • Note settings
    • Note Insights New
    • Engagement control
    • Make a copy
    • Transfer ownership
    • Delete this note
    • Save as template
    • Insert from template
    • Import from
      • Dropbox
      • Google Drive
      • Gist
      • Clipboard
    • Export to
      • Dropbox
      • Google Drive
      • Gist
    • Download
      • Markdown
      • HTML
      • Raw HTML
Menu Note settings Note Insights Versions and GitHub Sync Sharing URL Create Help
Create Create new note Create a note from template
Menu
Options
Engagement control Make a copy Transfer ownership Delete this note
Import from
Dropbox Google Drive Gist Clipboard
Export to
Dropbox Google Drive Gist
Download
Markdown HTML Raw HTML
Back
Sharing URL Link copied
/edit
View mode
  • Edit mode
  • View mode
  • Book mode
  • Slide mode
Edit mode View mode Book mode Slide mode
Customize slides
Note Permission
Read
Only me
  • Only me
  • Signed-in users
  • Everyone
Only me Signed-in users Everyone
Write
Only me
  • Only me
  • Signed-in users
  • Everyone
Only me Signed-in users Everyone
Engagement control Commenting, Suggest edit, Emoji Reply
  • Invite by email
    Invitee

    This note has no invitees

  • Publish Note

    Share your work with the world Congratulations! 🎉 Your note is out in the world Publish Note No publishing access yet

    Your note will be visible on your profile and discoverable by anyone.
    Your note is now live.
    This note is visible on your profile and discoverable online.
    Everyone on the web can find and read all notes of this public team.

    Your account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

    Your team account was recently created. Publishing will be available soon, allowing you to share notes on your public page and in search results.

    Explore these features while you wait
    Complete general settings
    Bookmark and like published notes
    Write a few more notes
    Complete general settings
    Write a few more notes
    See published notes
    Unpublish note
    Please check the box to agree to the Community Guidelines.
    View profile
    Engagement control
    Commenting
    Permission
    Disabled Forbidden Owners Signed-in users Everyone
    Enable
    Permission
    • Forbidden
    • Owners
    • Signed-in users
    • Everyone
    Suggest edit
    Permission
    Disabled Forbidden Owners Signed-in users Everyone
    Enable
    Permission
    • Forbidden
    • Owners
    • Signed-in users
    Emoji Reply
    Enable
    Import from Dropbox Google Drive Gist Clipboard
       Owned this note    Owned this note      
    Published Linked with GitHub
    • Any changes
      Be notified of any changes
    • Mention me
      Be notified of mention me
    • Unsubscribe
    ###### tags: `Paper`、`3D Computer Game`、`Vision-based Motion Capture` # **[Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition](https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/view/17135)** 本篇論文是 2018 年由 Sijie Yan, Yuanjun Xiong, Dahua Lin 發表於 AAAI 的 paper。 ## 1. Introduction 在之前的關於動態骨架識別的論文中,是在固定的一段時間取得關節坐標,並且作時間分析。然而這種方法並沒有利用到關節間的空間關係。 > 所謂步長就是你將需要觀測的數值均勻分成若干個區間,每個區間的長度就叫步長。 > > 時間步長法應該是以時間軸為主自變量,確定步長後定點測值,簡單來說就是每隔一段時間取一個值記錄下來。 > A feature vector is a specific observable phenomenon’s measurable property. > > 特徵向量是特定可觀察現象的可測量屬性。 人類類別中的身高和體重參數是一個明顯的例子,因為它是可觀察和可測量的。 <font color="red">由於骨架座標更適合用 graph 來表示,故這篇論文架構採用 graph neural network,並且延伸出一個名為 *Spatial-Temporal Graph Convolutional Network*(ST-GCN) 的 spatial temporal graph model,用以表示動作識別的骨架序列。</font> <center> ![The spatial temporal graph of a skeleton sequence.](https://i.imgur.com/MQz1Raf.png) </center> 如上圖,藍點為關節點,邊分為兩種,藍色為 spatial edges,綠色則為連接相同關節在不同時間的 temporal edges。這個模型的輸入值為關節坐標。 ## 2. Releted Work ## 3. Spatial Temporal Graph ConvNet ### 3.1 Pipeline Overview <center> ![Pipeline.](https://i.imgur.com/bI1Kmyn.png) <!--![Pipeline](https://i.imgur.com/15PnF0t.png)--> </center> 從 motion-capture 設備或是對影片進行姿勢預測(pose estimation) 取得骨架資訊,取得的數據是一系列的幀,每一幀都會有一組關節坐標,從這些 2D 或 3D 的坐標得出的骨架關節序列,並且建構出一個時空圖(spatial temporal graph) ,再將這些關節點的向量做為 input 輸入到 ST-GCN 模型中。輸入的數據經過運算後會在圖上形成 feature map,並且用 standard SoftMax classifier 分類到相應的動作類別。 <font color="red">簡單來說就是最終輸入到 ST-GCN model 的要是骨架資訊,且不限關節數量,而在這之前的處理方式可以有很多種,而這篇論文裡的 data set 有兩個 1) Kinetics 2) NTU-RGB+D,具體細節在後面會講。最終 model 會將輸入的 motion clip 分類到對應的 class。</font> ### 3.2 Skeleton Graph Construction 首先在每一幀取得每個關節點的坐標訊息,並且連接起來。而在本篇論文的工作,就是用時空圖將這些骨架序列進行分層表示。本篇論文在這個骨架序列上建構了一個含有 $N$ 個 關節點及 $T$ 幀的無向時空圖 $G = (V, E)$。 在此圖中,<font color="red">點集合可表示為 $V = \{v_{ti} \ | \ t = 1, ..., T, i = 1, ..., N \}$。</font>時空圖將分為兩個步驟建構出來,首先將相同幀內的所有關節點用邊,以符合人體結構的方式連接起來,然後將每個關節點與連續幀的同一關節點用邊連接起來。在這裡可以看到<font color="red">邊可分為兩個子集合,首先是連接同一幀的邊(intra-skeleton connection),可表示為</font> <font color="blue">$E_S = \{ v_{ti}v_{tj} \ | \ (i, j) \in H \}$</font>,<font color="red">另一個則是連接連續幀中的相同關節的邊(inter-skeleton),可表示為</font> <font color="green">$E_F = \{ v_{ti}v_{(t+1)i}\}$。</font> ### 3.3 Spatial Graph Convolutional Neural Network 論文這邊在看 ST_GCN 之前,先從單一幀的 graph CNN model 開始。在時間 $\tau$ 時,會有 $N$ 個關節點組成的集合 $V_t$,因此骨架的邊可以表示成 $E_s(\tau) = \{ v_{ti}v_{tj} \ | \ t = \tau, (i, j) \in H \}$,因為在單一幀,所以不會有 $E_F$ 這種邊。 ~~先從在單一幀的 graph CNN model,假設 kernel 大小為 $K \times K$,輸入的特徵圖為 $f_{in}$,且 channel 為 $c$,輸出的值在空間位置 $\mathbf{x}$ 為~~ 下面這行公式是一般 CNN convolution function,因為 GCN 就是 based on CNN,所以基本上就要先從這邊下手 $$f_{out}(\mathbf{x}) \ = \ \sum^K_{h=1} \sum^K_{w=1} {f_{in}(\mathbf{p}(\mathbf{x}, h, w)) \cdot \mathbf{w} (h,w)} \tag{1}$$ 可以看到這行公式分為 sampling 跟 weight 兩個部分 **sampling function $\mathbf{p}$:** $Z^2 \times Z^2 \to Z^2$ **weight function $\mathbf{w}$:** $Z^2 \to \Bbb{R}^c$ 再來將上述公式延伸到輸入特徵圖 spatial graph $V_t$,則特徵圖為 $f^t_{in} : V_t \to R^c$,接著重新定義 sampling function $\mathbf{p}$ 和 weight function $\mathbf{w}$。 **Sampling function:** 在 graph 中,對於 node $v_{ti}$,將其鄰近點的子集重新定義為 $B(v_{ti}) = \{v_{tj} | d(v_{tj}, v_{ti}) \leq D \}$,因此 sampling function $\mathbf{p} : B(v_{ti} \to V)$ 可以寫成 $$\mathbf{p}(v_{ti}, v_{tj}) = v_{tj} \tag{2}$$ <font color ="red">首先,先考慮在 image 上,image 會以 grid 的方式來考慮,因此以 $\mathbf{x}$ 為中心的相鄰的 (neighbor) 九宮格的點,都會被納入計算,將概念帶入到 graph 上,以 $v_{ti}$ 為中心,相鄰的點則可以定義為 $B(v_{ti}) = \{ v_{tj} \ | \ d(v_{tj}, v_{ti}) \leq D \}$,也就是說,兩點之間距離小於等於 $D$ 時,可以將 $v_{tj}$ 視為 $v_{ti}$ 的 neighbor,這裡的 $d(v_{tj}, v_{vi})$ 指的不是實際距離,而是 step,且在這篇論文中,$D$ 設為 1,也就是說在 one ring 內的點,皆為 neighbor。</font> **Weight function:** $l_{ti} : B(v_{ti}) \to \{ 0, ..., K - 1 \}$,weight function $\mathbf{w}(v_{ti}, v_{tj}): B(v_{ti}) \to R^c$ $$\mathbf{w}(v_{ti}, v_{tj}) = \mathbf{w}'(l_{ti}(v_{tj})) \tag{3}$$ <font color = "red">這邊作者考慮到不要給各個關節點固定的 labeling,而是以 $v_{ti}$ 的相鄰集劃分為 $K$ 個子集,因此每個子集皆有一個 label。label 細節在 Section 3.4。</font> **Spatial Graph Convolution.** 將公式 1,重新寫過後可得 $$f_{out}(v_{ti}) \ = \ \sum_{v_{tj} \in B(v_{ti})} {\frac{1}{Z_{ti}(v_{tj})} f_{in}(\mathbf{p}(v_{ti}, v_{tj})) \cdot \mathbf{w} (v_{ti}, v_{tj})} \tag{4}$$ 其中 normalizing term $Z_{ti} = |\{ v_{tk}|l_{ti}(v_{tj}) \}|$。 將公式 2 及公式 3 代入後可得 <font color = "red">spatial graph convolution 的最終公式: $$f_{out}(v_{ti}) \ = \ \sum_{v_{tj} \in B(v_{ti})} {\frac{1}{Z_{ti}(v_{tj})} f_{in}(v_{tj}) \cdot \mathbf{w} (l_{ti}(v_{tj}))} \tag{5}$$ </font> **Spatial Temporal Modeling.** 上面公式完成了 spatial 的部分,現在將 temporal 以同樣概念再帶入公式中,首先,sampling function 重新定義成 $$B(v_{ti}) = \{ v_{qj} = | d(v_{tj}, v_{ti}) \leq K, |q - t| \leq \lfloor \Gamma / 2 \rfloor \} \tag{6}$$ 其中 $\Gamma$ 控制幀數,而 weight function 則變成 label map $l_{ST}$ $$l_{ST}(v_{qi}) = l_{ti}(v_{tj}) + (q - t + \lfloor \Gamma / 2 \rfloor) \times K) \tag{7}$$ ### <font color = "red">3.4 Partition Strategies <center> ![](https://i.imgur.com/kU02uOc.png) </center> 對於動作識別,上圖展示了三種 labeling 規則: - **Uni-labeling:** 將所有相鄰的節點劃分為同一子集(綠色)。 - **Distance partitioning:** 具有兩個子集,root 本身為一子集(綠色),其餘相鄰的節點為同一子集(藍色)。 - **Spatial configuration partitioning:** 將該點本身集相鄰節點分為三種子集:1) root 本身,2) 比 root 更靠近骨架重心的相鄰節點,3) 較 root 遠離骨架重心的相鄰節點。 其中,第二種分法,考慮到了關節點之間的相對關係,而第三種則是考量到肌肉收縮與伸長對動作產生的不同影響。</font> ### 3.5 Learnable edge importance weighting ### 3.6 Implementing ST-GCN **Network architecture and training.** <center> ![Network architecture.](https://i.imgur.com/wNxvmvo.png) From [YouTube: Spatial-Temporal Graph Convolution Networks for skeleton-based Action Recognition](https://www.youtube.com/watch?v=JJZUvhIuvOQ) </center> ## 4. Experiments 在這篇論文中,作者用到了兩種 dataset,分別是 **Kenetics** 和 **NTU-RGB+D**,在 Section 4.1 中回詳細介紹。 ### 4.1 Dataset & Evaluation Metrics **Kenetics:** * 300,000 video clips * 400 human action classes * at least 600 video clips for each action class * Each clip of duration 10 seconds * Sampled at 30 fps * Feature vector per frame pre person - 18 nodes or joint locations - $(X, Y, C)$, where $(X, Y)$ are spatial coordinates and $C$ is xonfidence score - Multi-person cases <font color = "blue">這邊有一點比較要注意的是,這個 dataset 出來的只有影片沒有骨架資訊,所以若要套進 ST-GCN,則需先透過 [OpenPose](https://ieeexplore.ieee.org/document/8765346) 辨識出骨架資訊,大概是 18 個關節點。</font> **NTU-RGB+D:** * 56,000 action clips * 60 human action class * Two benchmark: 1) cross-subject (X-sub): - 40, 320 training - 16, 560 evaluation 2) cross-view (X-view): - 37, 920 training - 18, 960 evaluation * Kinect depth sensors - 3D joint locations $(X, Y, Z)$ <font color= "blue">這邊比較特別的部分是兩個 benchmark,一個是同一動作,不同人來做,另一個是同一個人,做不同動作。</font> ### 4.2 Ablation Study ### 4.3 Comparison with State of Arts 下表是作者以 Kinetics 為輸入,與其他論文做比較,可以看到,ST-GCN 相對其他論文,有更好的辨識準確率,且若增加 RGB 或 Optical Flow 資訊後,準確率可以有更大程度的提升。 <center> ![Kinetics.](https://i.imgur.com/XnxGFbB.png) </center> 下表是作者以 NTU-RGB+D 為輸入,與其他論文做比較後的結果,可以看到 ST-GCN 在此也有不錯的表現。 <center> ![NTU-RGB+D.](https://i.imgur.com/xR3tg1G.png) </center> ## 5. Conclusion 本篇論文的核心貢獻為有別於傳統骨架辨識時,僅依靠單一幀的空間資訊,作者將時間資訊同時也納入考量點,創造出名為 *Spatial-Temporal Graphic Neural Networks* (*ST-GCN*) 的新 model,顯示出若將時間資訊,也納入做為動作辨識的一項依據時,對於生成結果也較為理想。

    Import from clipboard

    Paste your markdown or webpage here...

    Advanced permission required

    Your current role can only read. Ask the system administrator to acquire write and comment permission.

    This team is disabled

    Sorry, this team is disabled. You can't edit this note.

    This note is locked

    Sorry, only owner can edit this note.

    Reach the limit

    Sorry, you've reached the max length this note can be.
    Please reduce the content or divide it to more notes, thank you!

    Import from Gist

    Import from Snippet

    or

    Export to Snippet

    Are you sure?

    Do you really want to delete this note?
    All users will lose their connection.

    Create a note from template

    Create a note from template

    Oops...
    This template has been removed or transferred.
    Upgrade
    All
    • All
    • Team
    No template.

    Create a template

    Upgrade

    Delete template

    Do you really want to delete this template?
    Turn this template into a regular note and keep its content, versions, and comments.

    This page need refresh

    You have an incompatible client version.
    Refresh to update.
    New version available!
    See releases notes here
    Refresh to enjoy new features.
    Your user state has changed.
    Refresh to load new user state.

    Sign in

    Forgot password
    or
    Sign in via Google Sign in via Facebook Sign in via X(Twitter) Sign in via GitHub Sign in via Dropbox Sign in with Wallet
    Wallet ( )
    Connect another wallet

    New to HackMD? Sign up

    By signing in, you agree to our terms of service.

    Help

    • English
    • 中文
    • Français
    • Deutsch
    • 日本語
    • Español
    • Català
    • Ελληνικά
    • Português
    • italiano
    • Türkçe
    • Русский
    • Nederlands
    • hrvatski jezik
    • język polski
    • Українська
    • हिन्दी
    • svenska
    • Esperanto
    • dansk

    Documents

    Help & Tutorial

    How to use Book mode

    Slide Example

    API Docs

    Edit in VSCode

    Install browser extension

    Contacts

    Feedback

    Discord

    Send us email

    Resources

    Releases

    Pricing

    Blog

    Policy

    Terms

    Privacy

    Cheatsheet

    Syntax Example Reference
    # Header Header 基本排版
    - Unordered List
    • Unordered List
    1. Ordered List
    1. Ordered List
    - [ ] Todo List
    • Todo List
    > Blockquote
    Blockquote
    **Bold font** Bold font
    *Italics font* Italics font
    ~~Strikethrough~~ Strikethrough
    19^th^ 19th
    H~2~O H2O
    ++Inserted text++ Inserted text
    ==Marked text== Marked text
    [link text](https:// "title") Link
    ![image alt](https:// "title") Image
    `Code` Code 在筆記中貼入程式碼
    ```javascript
    var i = 0;
    ```
    var i = 0;
    :smile: :smile: Emoji list
    {%youtube youtube_id %} Externals
    $L^aT_eX$ LaTeX
    :::info
    This is a alert area.
    :::

    This is a alert area.

    Versions and GitHub Sync
    Get Full History Access

    • Edit version name
    • Delete

    revision author avatar     named on  

    More Less

    Note content is identical to the latest version.
    Compare
      Choose a version
      No search result
      Version not found
    Sign in to link this note to GitHub
    Learn more
    This note is not linked with GitHub
     

    Feedback

    Submission failed, please try again

    Thanks for your support.

    On a scale of 0-10, how likely is it that you would recommend HackMD to your friends, family or business associates?

    Please give us some advice and help us improve HackMD.

     

    Thanks for your feedback

    Remove version name

    Do you want to remove this version name and description?

    Transfer ownership

    Transfer to
      Warning: is a public team. If you transfer note to this team, everyone on the web can find and read this note.

        Link with GitHub

        Please authorize HackMD on GitHub
        • Please sign in to GitHub and install the HackMD app on your GitHub repo.
        • HackMD links with GitHub through a GitHub App. You can choose which repo to install our App.
        Learn more  Sign in to GitHub

        Push the note to GitHub Push to GitHub Pull a file from GitHub

          Authorize again
         

        Choose which file to push to

        Select repo
        Refresh Authorize more repos
        Select branch
        Select file
        Select branch
        Choose version(s) to push
        • Save a new version and push
        • Choose from existing versions
        Include title and tags
        Available push count

        Pull from GitHub

         
        File from GitHub
        File from HackMD

        GitHub Link Settings

        File linked

        Linked by
        File path
        Last synced branch
        Available push count

        Danger Zone

        Unlink
        You will no longer receive notification when GitHub file changes after unlink.

        Syncing

        Push failed

        Push successfully