Dot 與 GraphViz 經驗總結

reference

• d3.js gallery: 僅僅是作個比較，d3的圖確實好看一點
• dot graphviz gallery: 其實也還不錯
• graphviz documence: 不同於官方文檔, github 上的文檔更新會快一點，當然個人覺得也好一點。
• dot pdf documence: 簡潔的pdf文檔，省去了多餘的東東。
• dot html documence: 很好的example
• dot 國人教程
• 高級API for style,type and other attributes
• color gallery

ubuntu

注：本人僅在ubuntu上進行了測試

安裝

• sudo apt-get install graphviz
  在ubuntu16.04上就這樣安裝好了，從源碼上安裝失敗，少了scan.c，反正能用，所以不管了。

使用

• dot -Tjpg example.dot -oexample.jpg
• example.dot

graph graphname{
a--b;
b--c;
b--d;
d--a;
}

• 指明輸出格式爲jpg, -Tjpg, 所有輸出格式
• 指明輸出文件名爲example.jpg -oexample.jpg
• example.jpg
  

An Ugly Tool and A nice Plugin

• DotEditor: 編輯功能極弱，用法是用別的編輯器進行dot文件編輯，再複製到DotEditor中進行顯示。
• [gedit plugin external tool]：自定義功能極強，可結合gedit 的編輯能力，非常好用。在下面的腳本和圖示配置下，按F5即可顯示dot圖像。

#!/bin/sh
format=jpg
fin=${GEDIT_CURRENT_DOCUMENT_NAME}
fout=${GEDIT_CURRENT_DOCUMENT_NAME}.${format}
touch ${fout}
echo ${fin}
echo ${fout}
dot -T${format} $fin -o${fout} & gnome-open ${fout}

• 更多gedit變量及external tool的用法可以參考官網
• 更多的gedit external tool 插件腳本

---

High Level Summary

• for subgraph, the cluster_i(i=0,1,2,…) is special
• we can use {...} to set local environment
• set node node [style=filled,color=lightpink];
• set edge a->b [label='...']
• use set {a,b,c}->c; a->{b,c,d}
• set edge direction edge [dir=both/none/forward/back;penwidth=3] , digraph,graph
• set graph direction rankdir="TB", "LR", "BT", "RL"
• for dash,solid... see style at 高級API for style,type and other attributes
• for color node [style=filled,color=lightyellow]; see color gallery

digraph G {

    subgraph Ncluster_0 {
        style=filled;
        color=lightgrey;
        node [style=filled,color=lightgray];
        start -> "motion detection" -> "connected component analysis" -> "small area filter" -> "object feature extraction" -> "object association" -> "association graph" -> "find min tree";
        class -> "IBGS.process()" -> "icvprCcaByTwoPass()" -> "QYzbxTrackingFeature.getObjectsFromMask()" -> "QYzbxTrackingFeature.getObjects()" -> "TrackingObjectAssociation.adjecentObjectAssociation()" -> "TrackingObjectAssociation.process()" -> "MotionBasedTracker.objectTracking()";
        data -> "image:CV_8UC3" -> "mask:CV_8UC1" -> "vector<mask>" -> "vector<ObjectFeature>" -> "list<FrameFeature>" ->"DirectGraph"->"tracking record";
        label="2016/08/20: Connected Component Analysis"
    }

    subgraph Ncluster_1 {
        style=filled
        color=blue
        node [style=filled,color=lightpink];
        "motion detection" -> "contour analysis" -> "kalman filter" -> "hungrian algorithm" -> "trakcking result"
        data -> "image:CV_8UC3" -> "mask:CV_8UC1" -> "contour:vector<Point>" -> "vector<trackingObjectFeature>" -> "MultiObjectTracker:vector<singleObjectTracker>"
        label="2016/08/22: findContour"
    }

    subgraph Ncluster_2 {
        style=filled
        color=red
        node [style=filled,color=lightyellow];
        "motion detection" -> "blob detection" -> "kalman filter" -> "hungrian algorithm" -> "trakcking result"
        data -> "image:CV_8UC3" -> "mask:CV_8UC1" -> "blob feature" -> "vector<trackingObjectFeature>" -> "MultiObjectTracker:vector<singleObjectTracker>"
        label="2016/08/23: blob detector"
    }

}

digraph G{
    {node[style=filled,color=lightyellow]
        input
        "QString:configFile"
        "QString:videoFile"
        "TrackingStatus*:status"
    }

    {node[style=filled,color=lightpink]
        function
        "void:process()"
        "void:processOne()"
        "vector<trackingObjectFeature>:getObjects()"
        "void:Tracking()"
        "void:run()"
        "void:stop()"
    }

    {node[style=filled,color=lightblue]
        explain
        "show tracking result"
        "init singleObjectTracker"
        "update singleObjectTracker"
    }

    {"QString:configFile","QString:videoFile"}->"void:process()"
    "void:process()"->{"Mat:img_input","Mat:img_foreground","Mat:img_background"}
    "void:process()"->"void:processOne()"->"vector<trackingObjectFeature>:getObjects()"->"vector<trackingObjectFeature>"->"void:Tracking()"
    "void:Tracking()"->"init singleObjectTracker"->"update singleObjectTracker"[label="kalman and hungrian"]
    "void:processOne()"->"show tracking result"
    "void:process()"->"void:run()"

    {"Mat:img_input","Mat:img_foreground","Mat:img_background","TrackingStatus*:status"}->"void:processOne()"

    {edge [dir=both;penwidth=3]
        "TrackingStatus*:status"->"TrackingStatus*:globalTrackingStatus"[label="always equal to"]
        {"TrackingStatus*:globalTrackingStatus","bool:globalStop"}->"void:run()"
        "void:process()"->"TrackingStatus*:status"
    }

    "void:stop()"->"bool:globalStop"
}