Implement dropblock_2d regularization layer in keras 3

[divyashreepathihalli]

keras-cv/keras_cv/layers/regularization/dropblock_2d.py

[sami-bg]

Do you mind if I give this a shot? It will be my first issue; I will be transliterating the layer's tf ops into the relevant keras ops. I will also write tests to make sure both implementations (keras 2 using TF and keras 3 using keras ops) match.

[divyashreepathihalli]

@sami-bg thank you for your interest. This layer would not be straightforward because it is subclassing an Keras 2 legacy internal baseRandomLayer. It could be a complicated issue for the first time.

[sami-bg]

Oh I see, I'll try for another issue then. Thanks for letting me know!

[divyashreepathihalli]

No worries! If you are interested we would love to have as many computer vision keras io examples converted to keras 3. This is a priority and could be a good one to start with. - https://github.com/keras-team/keras-io/tree/master/examples/vision

here is a list of them we are yet to convert

• [ ] learnable_resizer
• [ ] deit
• [ ] mixup
• [ ] perceiver_image_classification
• [ ] shiftvit
• [ ] bit
• [ ] simsiam
• [ ] generative/dcgan_overriding_train_ste
• [ ] forwardforward
• [ ] nnclr
• [ ] handwriting_recognition
• [ ] xray_classification_with_tpus
• [ ] edsr
• [ ] mobilevit
• [ ] yolov8
• [ ] video_classification
• [ ] nl_image_search
• [ ] supervised-contrastive-learning
• [ ] depth_estimation
• [ ] fully_convolutional_network
• [ ] vivit
• [ ] randaugment
• [ ] consistency_training
• [ ] basnet_segmentation
• [ ] barlow_twins
• [ ] adamatch
• [ ] pointnet
• [ ] vit_small_ds
• [ ] retinanet
• [ ] temporal_latent_bottleneck
• [ ] near_dup_search
• [ ] 3D_image_classification
• [ ] swin_transformers
• [ ] focal_modulation_network

[sami-bg]

I'll give it a try! Thank you for the suggestion :)