Object detection on multiple datasets with an automatically learned unified label space.

A package to read and convert object detection datasets (COCO, YOLO, PascalVOC, LabelMe, CVAT, OpenImage, ...) and evaluate them with COCO and PascalVOC metrics.

Scaling Object Detection by Transferring Classification Weights

Convert OpenImages labels to be used for YOLOv3

Open Images is a dataset of ~9M images annotated with image-level labels, object bounding boxes, object segmentation masks, visual relationships, and localized narratives.

Convert openimages v4 dataset to darknet train datas.

Inceptionv3 model trained on OpenImages dataset, with code in Tensorflow

Mahajan et al. and Yalniz et al. demonstrate the benefit of pre-training on large unlabeled image datasets using user tags as labels for weak-supervision. We take this approach further, considering whether the use of user tags in large labeled image datasets is beneficial.

Download sub part of open image dataset

Tools developed for sampling and downloading subsets of Open Images V5 dataset and joining it with YFCC100M

I improved the original toolkit for downloading images using OpenAI images datasets - OpenImages Downloader to add Resumable and version changing capabilities. This toolkit also supports xml as well as txt files as input and output.