Maxim Zhiltsov

Requirements: - more dedicated support of track annotations - ReID annotations - images as annotations (image-to-image tasks - super resolution, style transfer, image generation, etc.) - (depth-)field annotations (distance prediction...

Loading of plugins can be expensive, as it actually is - for TF and Accuracy Checker plugins, which can be loaded in few seconds. Plugins need to be reorganized in...

Currently, all annotations can have arbitrary attributes, but there is no specifications for these attributes. It would be great, if Datumaro provided: - A list of possible attribute values (at...

A single PDF can be considered as a container with several images, each of which can be used for annotation. Look at [CVAT implementation of PDF data](https://github.com/openvinotoolkit/cvat/pull/2154) as a reference.

Once a dataset is located or imported, it can be useful to provide source file information and other format-related properties, such as image directory, source files, presence of images etc....

A video can contain lots of frames, most of which do not bring any new information. A dataset created from a video can be significantly reduced without affecting model performance,...

Datasets often come as zip or tar archives, it would be nice to support such kind of inputs in Datumaro. It can be undesirable to unpack them, because this can...

Command-line interface is one of the 2 main interfaces for the project. Currently, the only tested part of Datumaro - is its model, the library API, along with most of...

Currently, to add a new format, an `Extractor` plugin is needed, which can be accompanied by an `Importer`. In the essence, these 2 entities serve a single purpose - importing...

Clouds and other kinds of remotes, which include s3, Azure, Google Drive, NFS & CIFS file shares, Git, Git-lfs, etc. are common places for storing large datasets. Support of such...