Add a caption_id to coco_gen dataset - Githubissues

mlcommons / mobile_app_open

Mobile App Open

https://mlcommons.org/en/groups/inference-mobile/

Apache License 2.0

47 stars 23 forks source link

Add a caption_id to coco_gen dataset #918

Closed anhappdev closed 2 months ago

anhappdev commented 2 months ago

I added a caption_id to the *.tfrecord files. The caption_id is the id field in the captions_source.tsv.
The caption_id will be used as the filename for the image generated by the mobile app.
With a caption_id in filename we can calculate any metrics (e.g. IQA-A) offline and map it with the spreadsheet for analytic purpose.

Relevant files are here: https://github.com/anhappdev/tmp/releases/tag/6

github-actions[bot] commented 2 months ago

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

sonarcloud[bot] commented 2 months ago

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarCloud

freedomtan commented 2 months ago

@mohitmundhragithub to test it and report if it's OK.

mohitmundhragithub commented 2 months ago

This PR seems to be working fine.

@anhappdev do you want to udpate the link for the dataset for the test dataset as well?