Disable/customise augmentation for the object_detector task

[DoctorDinosaur]

MediaPipe Solution (you are using)

Object Detection (Model Maker)

Programming language

Python

Are you willing to contribute it

No

Describe the feature and the current behaviour/state

Currently, while the Image Classification task has "do_data_augmentation", the Object Detection Task does not. The object detection task should be provided this option.

It would also be useful if both could allow for customisation. TF-Vision exposes aug_scale_min, aug_scale_max, aug_rand_hflip, aug_rand_vflip, and aug_type; where aug_type can take a tfm.vision.configs.common.Augmentation (i.e. AutoAugment or RandAugment)

Will this change the current API? How?

No response

Who will benefit with this feature?

No response

Please specify the use cases for this feature

My dataset includes "right arrow" and "left arrow". Random Flipping is enforced and cannot be disabled, without subclassing the object detection class. This means the classes cannot be trained.

Any Other info

No response

[kuaashish]

Hi @joezoug,

Do we have a plan to incorporate this feature into the Object Detection model maker?

Thank you!!

[joezoug]

Hi @DoctorDinosaur,

Thanks for the suggestion. Unfortunately the team doesn't have bandwidth to make this change in the near future. Can you try subclassing/overriding the class to disable data augmentation?

[github-actions[bot]]

This issue has been marked stale because it has no recent activity since 7 days. It will be closed if no further activity occurs. Thank you.

[github-actions[bot]]

This issue was closed due to lack of activity after being marked stale for past 7 days.