Hsinchu city traffic image data_with point center annotation_5000p

Dataset Description

• This data set is a fisheye lens video file of important intersections in Hsinchu City, provided by Hsinchu City Police Department. Original video reference:

  • (This video is from the fish-eye in front of Hsinchu Railway Station on Linsen Road 2019/11/15) https://youtu.be/S-tYTAaQgIs

• Contains traffic data sets of four intersections in Hsinchu City, including about 500 images and about 5,000 point annotations

  • Use the point center labeling method,
  • Only objects within the ROI range are marked
  • The content of the label is divided into: large car, small car, locomotive

• It can be used for AI scientific research and development, everyone is welcome to apply.

Data information

• This information is provided by Hsinchu City Police Department
• Marked by Ruoshui Technology
• And the data service is provided by the State Grid Center

video date

• November 2019

file description

• This data set contains four compressed files. After decompression, the folders represent each intersection.
  • 01.zip => Guangfu Road
  • 09.zip => Jinguo Road
  • 10.zip => Railway Station
  • 12.zip => Jieshou Road

Each compressed file contains an array of png images, json annotation files, and bound annotation images. The following examples illustrate:

• 01_roi.txt => mark the roi range
• frame_00000.png => original image
• frame_00000.json => annotation file

• frame_00000_out.png => Binding annotated image

• Figure 1: The marked frame_00000.png, red dots are locomotives, green dots are small cars, blue dots are large cars, the ROI range is near the intersection area within the zebra crossing, please refer to the roi.txt of each intersection for details

• Figure 2: roi.txt two points are combined into a line, there are multiple lines at one intersection

  • line : [ [x1,y1], [x2,y2] ]

• Figure 3: Mark json file reference

Demonstration application

• ai automatic labeling application (this video is the former fisheye of Hsinchu Industrial Research Institute, using AI model to automatically determine 2021/09/27) https://youtu.be/sjUQ-Ayxxtk

資料集說明

• 此資料集為新竹市重要路口的魚眼鏡頭影片檔，由新竹市警察局提供。原始影片參考：

  • (此段影片為 新竹火車站前魚眼照林森路 2019/11/15) https://youtu.be/S-tYTAaQgIs

• 內含四個新竹市路口的交通資料集，含約500張影像與約5000個點標註

  • 使用 點中心 標註方式，
  • 在ROI範圍內的物件才標註
  • 標註內容分為: 大型車、小型車、機車

• 可用於AI科學研發，歡迎大家申請使用。

資料資訊

• 此資料由新竹市警察局提供
• 由若水科技標註
• 並由國網中心提供資料服務

影片日期

• 2019 年 11 月

檔案說明

• 此資料集包含四個壓縮檔，解壓縮後資料夾分別代表各個路口
  • 01.zip => 光復路
  • 09.zip => 經國路
  • 10.zip => 火車站
  • 12.zip => 介壽路

每個壓縮檔內包含數組 png影像、json標註檔、綁定標註的影像。 以下舉例說明:

• 01_roi.txt => 標註 roi 範圍
• frame_00000.png => 原始影像
• frame_00000.json => 標註檔

• frame_00000_out.png => 綁定標註的影像

• 圖一：經標註的frame_00000.png，紅點為機車，綠點為小型車，藍點為大型車，ROI範圍在斑馬線內的十字路口區附近，詳細參考各路口的 roi.txt

• 圖二：roi.txt 兩點組合成一條線，一個路口有多條線

  • line : [ [x1,y1], [x2,y2] ]

• 圖三 : 標註json file 參考

示範應用

• ai 自動標註應用 (此段影片為 新竹工研院前魚眼，使用AI模型自動判定 2021/09/27) https://youtu.be/sjUQ-Ayxxtk