Box2Seg
Box2Seg is a semantic annotation tool but faster.
Why another annotation tool
Deep learning-based computer vision models are becoming more prominent with various applications in medical imaging, automotive, logistics, etc. However, they require a large amount of annotated data to perform significantly well, and generating annotation labels for these large datasets is a cumbersome and time-consuming task. Specifically, for segmentation applications generating annotated data is a relatively difficult task compared to drawing simple bounding boxes for object detection. For example, bounding box labels require just two points, whereas semantic labels require a minimum of 20 points, even for a coarse label, as shown in Figure 1 (manually generated using labelme annotation tool1)
References
-
G. Bradski. The OpenCV Library. Dr. Dobb’s Journal of Software Tools, 2000.
-
Robin Del´earde, Camille Kurtz, Philippe Dejean, and Laurent Wendling. Segment my object: A pipeline to extract segmented objects in images based on labels or bounding boxes. In VISIGRAPP (5: VISAPP), pages 618–625, 2021.
-
Martin Rajchl, Matthew CH Lee, Ozan Oktay, Konstantinos Kamnitsas, Jonathan Passerat- Palmbach, Wenjia Bai, Mellisa Damodaram, Mary A Rutherford, Joseph V Hajnal, Bernhard Kainz, et al. Deepcut: Object segmentation from bounding box annotations using convolutional neural networks. IEEE transactions on medical imaging, 36(2):674–683, 2016.
-
Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian Reid, and Silvio Savarese. Generalized intersection over union: A metric and a loss for bounding box regression. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 658–666, 2019.
-
Carsten Rother, Vladimir Kolmogorov, and Andrew Blake. ”GrabCut”. ACM Transactions on Graphics, 23(3):309–314, August 2004. doi: 10.1145/1015706.1015720. URL https://doi.org/10.1145/1015706.1015720.