Page MenuHomePhabricator

Test the feasibility of a classifier trained on Commons categories
Open, Needs TriagePublic

Description

  • Identify meaningful categories/groups of categories whose images can be used for training T242969
  • Download images for categories
  • Test accuracy and performance of image classifiers (pretrained or trained from scratch) for various sets of categories T242970
  • Report on results in terms of accuracy and GPU performances T242970
  • [Optional] test different DL architectures

Event Timeline

Miriam moved this task from Staged to In Progress on the Research board.Jan 8 2020, 4:28 PM

I'm just subscribing because this sounds super interesting ;)

Weekly update:

  • downloaded the list of coco-stuff classes which include highly generic categories of people, animals, and things which exist in the visual world: https://github.com/nightrome/cocostuff
  • downloaded the list of categories in Commons, with the counts of the number of images per categories.
  • to create the initial seed of categories we want to consider for object categorization in Commons, I computed fasttext vectors on both COCO categories and Commons Categories, and I am checking what are the commons categories that we can use to represent COCO categories.
Miriam updated the task description. (Show Details)Jan 16 2020, 1:36 PM
Miriam removed subscribers: Mholloway, Cirdan.