Elie Bursztein

Implement DINO

5

https://github.com/facebookresearch/dino?fbclid=IwAR0gexQPdJdMe_ZyXqfWSDOFvLG1TcrVZELOb3ihvBYR262UTMpehRL-LZo

Figure out how to deal with large-scale dataset

1

We can do our own using a gpu accelerate lib - https://medium.com/rapids-ai/scaling-knn-to-new-heights-using-rapids-cuml-and-dask-63410983acfe - https://github.com/facebookresearch/faiss - https://github.com/yahoojapan/NGT (vlad algorithm) # serving - https://oss.redislabs.com/redisai/intro/ or use a fully fledge server - https://github.com/vdaas/vald...

Implement multi-siam for segmentation

5

https://arxiv.org/pdf/2108.12178.pdf

Add a ranking example

3

See for example: http://surpriselib.com/ https://github.com/tensorflow/ranking

Lifted Structured loss

9

Implement the Lifted Structured Loss - https://arxiv.org/abs/1511.06452 https://www.tensorflow.org/addons/api_docs/python/tfa/losses/LiftedStructLoss

Add from_config to ensure serialization

1

https://github.com/keras-team/keras-io/pull/628/files

Port the gradcam++ code