[data request] Pascal VOC

[rodrigob]

• Name of dataset: Pascal VOC12 with SDB extension
• URL of dataset: http://host.robots.ox.ac.uk/pascal/VOC/ and http://home.bharathh.info/pubs/codes/SBD/download.html
• License of dataset: ?? (to be investigated)
• Short description of dataset and use case(s): the classic detection and semantic segmentation dataset. Used in many works related to weakly supervised learning (since a bit easier than Coco).

Folks who would also like to see this dataset in tensorflow/datasets, please +1/thumbs-up so the developers can know which requests to prioritize.

[ParthS007]

Hi @rsepassi, I would like to work on adding this dataset. I have gone though gitter discussions and want to get started with this Project. Can you please help me getting started and assign this issue to me.

Thanks

[rsepassi]

Great! Let me know when you've accepted the collaborator invite, and I'll assign you. The guide for adding a new dataset is here: https://github.com/tensorflow/datasets/blob/master/docs/add_dataset.md

On Sat, Mar 2, 2019 at 11:03 AM Parth Shandilya notifications@github.com wrote:

Hi @rsepassi https://github.com/rsepassi, I would like to work on adding this dataset. I have gone though gitter discussions and want to get started with this Project. Can you please help me getting started and assign this issue to me.

Thanks

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/tensorflow/datasets/issues/27#issuecomment-468950002, or mute the thread https://github.com/notifications/unsubscribe-auth/ABEGW6AHCDanVm7WLjiS9RRrcsBKe7YTks5vSssEgaJpZM4aNAdD .

[ParthS007]

@rsepassi I have accepted the invite. I will go through the doc and will ask here if I face any difficulty.

Thanks for the help.

[ParthS007]

@rodrigob Can you please provide the exact links of training and validation datasets?

[thealphacod3r]

@rodrigob I would love to work on this dataset if anyone not working on it can you assign it to me??

[ParthS007]

@amanp592 You can work on it.

[Conchylicultor]

@ParthS007 @amanp592 Please note that someone already started working on this: https://github.com/tensorflow/datasets/pull/343

[jpuigcerver]

Hi,

I'm the author of #343: JFYI my pull request only adds VOC2007, not VOC2012.

[Conchylicultor]

@jpuigcerver oops, my bad, indeed. Sorry about that.

[rodrigob]

@ParthS007 Any progress on this front ?

[AdamHillier]

I needed VOC2012 for something so added it to my fork of tfds. It is simply @jpuigcerver's VOC2007 code adapted very slightly to point to the right tar files.

I've opened a PR #966 to have it merged.

As it includes only the object detection ground truths (axis aligned bounding boxes) and not the segmentation ground truths (or the SDB extension) it doesn't fully satisfy this dataset request, but it might be useful to some people.

(Note that it includes only the training and validation VOC2012 data, as annotations for the test data are not publicly available.)

[Conchylicultor]

voc/2012 is now available. Closing this

[rodrigob]

This ticket was not understood. https://www.tensorflow.org/datasets/catalog/voc only has bounding boxes, while Pascal VOC12 with SDB extension includes semantic labeling annotations (as indicated in the ticket) which are still missing in tfds. It seems that semantic/instance/panoptic segmentation are missing as a category of image annotations.

I would argue the ticket should be re-opened (maybe title improved).

[Conchylicultor]

@rodrigob thanks for the clarification. Re-opening.

[AdamHillier]

This ticket was not understood. https://www.tensorflow.org/datasets/catalog/voc only has bounding boxes, while Pascal VOC12 with SDB extension includes semantic labeling annotations (as indicated in the ticket) which are still missing in tfds. It seems that semantic/instance/panoptic segmentation are missing as a category of image annotations.

I have PR #1198 open to add VOC segmentation with the SBD extension.

[bhack]

For the segmentation take a look at https://github.com/keras-team/keras-cv/pull/943