This repository contains:
1 A software to label image
2 APIs convert detection data to other formats(VOC,COCO,YOLO)
In the released folder,there are two versions for Windows7 and Mac.
No library installation requirements.
This tool can be used for making classification, tracking, and detection Datasets.
After type the username and password, click Classification button
Click Open Dir Button to select a image
Choose a checkbox,and click classify button. You can also go to any frame with the GO button on the upright corner.
The classied images will be copied to corresponding folders.
Click Tracking Button, and draw bounding box by mouse, the coordinates of bounding box will be shown in the bottom.
Click track button, then go to next image. The results will be saved in the results.txt
The format is : image_name x/width y/height w/width h/height
Click Detection Button, draw bouding box of one object, then choose a checkbox and click detect object button. After drawn all objects in this frame, click detect image button and go to next image.
The results will be saved in Detection/image_name.txt.
The format is : label x/width y/height w/width h/height
Different detection platforms require different formats of data.So it is impossible to use our detection dataset from the tool directly.I will provide APIs to convert our format to popular detection formats easily. Right now, it only surport VOC, COCO, YOLO. I will consider to add tensorflows' format and Mxnets'format.
opencv3
lxml
xml
python main.py
--detected_dir "/home/Detection"
--image_path "/home/Images"
--saved_dir "/home/VOC2007"
--train_percent 0.9
--convert_to voc
detected_dir: the path of detection data made from this tool
image_path : the path of original images
saved_dir : the directory you want to save the converted data
train_percent : how many percent of image you want to train
convert_to: which format you want to convert to (choose one from "voc","yolo", "coco")