[go: up one dir, main page]
Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add converters for VIAME CSV <-> COCO #101

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Add converters for VIAME CSV <-> COCO #101

wants to merge 1 commit into from

Conversation

predicative
Copy link
Collaborator

This PR (currently) simply adds two pipeline files to convert VIAME CSV to and from the COCO format to the detection_file_conversions example.

Some possible points of attention:

  • The other pipelines in this folder use an image list as input; these don't. (I'm not sure of the advantage of the former approach.)
  • Related to the previous point, the CSV -> COCO converter requires Teach read_detected_object_set_viame_csv to return file names #99 to work properly.
  • I have a separate Python script, not included, to extract an image list from a COCO file, which is useful for creating an image list to accompany the CSV resulting from conversion. Do we also want this? If so, where? I have another script that runs both the conversion and the image list extraction, and again the question is where to put it.
  • Is there a summary of the COCO format I could reference for this example's README? The format used does depart slightly from the one at http://cocodataset.org/#format-data, namely in the addition of a "score" attribute currently, but perhaps supporting more of the Kitware extensions later.

@mattdawkins
Copy link
Collaborator

The image list input is to ensure that the frame_ids in the csv (if used) match the position in the input list in case a custom ordering of input images is desired

I think you want to refer to this as a KWIVER .json format in the readme, but still say that it still supports native COCO (and link to it), as we're likely to add in tracking support if using it and brand it as that.

@mattdawkins
Copy link
Collaborator
mattdawkins commented Feb 13, 2020 •
edited
Loading

w.r.t. the whole python binding / coco implementation setup it would be nice if the coco readers could support this (what the other ones do is if a file string (path) is passed into read_set, then it produces detections for that file, if it's blank it reads it in the order as stored in the original file)

@predicative predicative mentioned this pull request Feb 19, 2020
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mattdawkins mattdawkins Awaiting requested review from mattdawkins

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@predicative @mattdawkins