Availability of training data in non-consumer visual domain determines how good the modeling pipeline is. In this project, we propose an interative computer vision algorithm (based on histogram inference) that in each iteration refines segmentation masks from existing annotation boxes in maritime domain.
A web wrapper for the Boxes2Masks application.
Aerial Image processing pipeline using Tensorflow implementation of MaskRCNN.
Deep Learning Framework is sensitive to noisy annotations. This tool extends VIA for COCO JSON compatibility, and allows for labeling on multiple levels of description; it allows Import, Export, and Edit Video Annotations in COCO JSON format.