Microsoft's Visual Object Tagging Tool (VoTT) is a popular annotation tool open sourced in 2017. It's a good choice for DIY labeling and we have a full VoTT tutorial to get you started using VoTT to annotate your own object detection dataset.
Aside from its JSON export format, VoTT also has a proprietary CSV format that is not supported by any known models. To train a model with your VoTT CSV you can simply use Roboflow to convert them to the format your machine learning model needs.
Below, learn the structure of VoTT CSV.
"image","xmin","ymin","xmax","ymax","label"
"img0001.jpg",109.02857142857141,86.14285714285714,153.77142857142854,123.94285714285714,"helmet"
"img0002.jpg",122.69760696156635,18.85103626943005,193.18346627991298,88.48834196891191,"person"
"img0003.jpg",6.816997518610422,22.483428571428572,195.0452853598015,182.48685714285713,"helmet"
"img0004.jpg",94.72403754752852,51.239999999999995,186.11187024714832,120.04799999999999,"helmet"
"img0005.jpg",67.98611408390411,14.411428571428573,183.07258668664383,108.64,"person"
"img0006.jpg",28.285714285714285,7.971428571428571,137.57142857142856,119.05714285714285,"helmet"
"img0007.jpg",66.6,55.028571428571425,195.42857142857142,193.37142857142857,"helmet"
"img0008.jpg",22.97155928938356,8.203428571428573,210.3474154537671,181.14057142857143,"helmet"
"img0009.jpg",47.20812182741117,4.705958549222798,251.26903553299493,168,"helmet"
"img0010.jpg",74.2909429280397,12.224,180.6333746898263,97.57371428571429,"helmet"With Roboflow supervision, an open source Python package with utilities for completing computer vision tasks, you can merge and split detections in VoTT CSV. Read our dedicated guides to learn how to merge and split VoTT CSV detections.
Below, see model architectures that require data in the VoTT CSV format when training a new model.
On each page below, you can find links to our guides that show how to plot predictions from the model, and complete other common tasks like detecting small objects with the model.