polimorfo 0.8.4

a dataset loader and converter for object detection segmentation and classification

polimòrfo

Polimòrfo (πολύμορϕος, comp. di πολυ- «poli-» e μορϕή «forma») is a dataset loader and converter library for object detection segmentation and classification. The goal of the project is to create a library able to process dataset in format:

• COCO: Common Objects in Context

• Pascal VOC: Visual Object Classes Challenge

• Google Open Images: Object Detection and Segmentation dataset released by Google

and transform these dataset into a common format (COCO).

Moreover, the library offers utilies to handle (load, convert, store and transform) the various type of annotations. This is important when you need to: - convert mask to polygons - store mask in a efficient format - convert mask/poygons into bounding boxes

• Free software: Apache Software License 2.0

• Documentation: https://polimorfo.readthedocs.io.

Features

TODO

• [X] Coco dataset

• [X] download coco datasets for train and val

• [X] add annotations loader and converter

• [X] add the ability to create dataet from scratch

• [ ] add voc dataset format

• [ ]

Credits

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.

History

0.2.0 (2020-02-18)

• Add support to process coco dataset

0.2.1 (2020-02-28)

• add support to download files and archives from the web and google drive

0.3.0 (2020-10-04)

• addedd support for removing categories and other utilities

0.4.0 (2020-10-05)

• addedd support to create a dataset from scratch

0.5.0 (2020-10-06)

• added support to visualize images and annotations

• make image removing optional during annotations and categories deletion

0.6.0 (2020-10-12)

• added copy dataset

• added split dataset

0.6.1 (2020-10-12)

• fixed a bug in colors generation for show images

0.6.2 (2020-10-12)

• update signature for function def update_images_path(self, func):

0.7.0 (2020-10-19)

• add method to dump dataset in format segmentation map

0.8.0 (2020-10-23)

• fixed bug in maskutils.mask_to_polygons

• add class to transform the predictions from instance and semantic segmentation in coco format

• fixed bug in add_image, add_annotation, add_category

• make load_image and load_images load random images sampled from the dataset

0.8.1 (2020-10-23)

• fixed bug for tqdm when removing a category and its annotations from the dataset

0.8.2 (2020-10-23)

• removed the prefix jpg when saving masks

• update draw instance to draw only bounding boxes

0.8.3 (2020-10-23)

• fixed bug in enum for draw instances

0.8.4 (2020-10-23)

• add show bounding boxes