Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

TAPVid-3D v1.0: queries_xyt (x, y) does not match query frame t

Open
  • Is something wrong?
#171 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
48/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
jupyter-notebook
Domain
data

Research direction

Start by comparing the v1.0 and rc5 annotations for the example clip Lite_release_recognition_BookDeepLearning_seq031_3, track 7 (rc5 track 10), and check how query frame t and coordinates (x, y) are represented. The issue lists inconsistent-track counts for 22 ADT clips and notes that other ADT clips, PStudio, and DriveTrack are unchecked. Done means correcting the affected query coordinates and checking the remaining annotation sets for the same mismatch.

Written by the indexing model from the issue text.

Description

In the v1.0 ADT annotations, some queries have an (x, y) that does not match the track's projection at the query frame t. For these tracks, (x, y) matches the projection at the rc5 query frame instead, and that frame is GT-occluded. It looks like t was moved to a visible frame after #131, but (x, y) was not updated.

Image

Example: Lite_release_recognition_BookDeepLearning_seq031_3, v1.0 track 7 (rc5 track 10).
Left: at the v1.0 query frame t = 155, the query (red ×) is 96 px from the track (green ○).
Middle: at the rc5 query frame t = 44, the same (x, y) is on the track, but the point is GT-occluded.
Right: only t changed between rc5 and v1.0.

clip inconsistent / tracks
Apartment_release_decoration_seq138_5 1 / 960
Apartment_release_decoration_seq138_6 0 / 873
Apartment_release_meal_seq136_2 26 / 707
Apartment_release_meal_seq136_8 31 / 695
Apartment_release_work_seq108_5 33 / 791
Lite_release_recognition_BambooPlate_seq030_0 20 / 943
Lite_release_recognition_BambooPlate_seq030_5 45 / 879
Lite_release_recognition_BambooPlate_seq030_6 30 / 805
Lite_release_recognition_BambooPlate_seq031_9 68 / 746
Lite_release_recognition_BirdHouseToy_seq030_6 40 / 873
Lite_release_recognition_BirdHouseToy_seq033_2 35 / 805
Lite_release_recognition_BookDeepLearning_seq030_0 43 / 855
Lite_release_recognition_BookDeepLearning_seq030_4 47 / 666
Lite_release_recognition_BookDeepLearning_seq030_5 65 / 838
Lite_release_recognition_BookDeepLearning_seq031_3 47 / 623
Lite_release_recognition_ChoppingBoard_seq031_0 8 / 987
Lite_release_recognition_ChoppingBoard_seq031_3 8 / 867
Lite_release_recognition_ChoppingBoard_seq033_1 57 / 646
Lite_release_recognition_DinoToy_seq030_0 41 / 894
Lite_release_recognition_Flask_seq030_0 34 / 906
Lite_release_recognition_Flask_seq030_3 37 / 915
Lite_release_recognition_WoodenBowl_seq032_1 17 / 502

I only checked these 22 ADT clips. I have not checked the other ADT clips, PStudio, or DriveTrack.

Dominant language
Jupyter Notebook
Stars
2.2k
Forks
215
No merged PRs in 30d

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from google-deepmind/tapnet

All issues in google-deepmind/tapnet

Similar issues

More Data Engineering issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.