rajat koner

intern @DeepMind, Ph.D. student @LMU_Muenchen with @vtresp,
Informatics @TU_Muenchen ,

[rajat koner @rajatkoner]

This idea inherently applicable to Image/Video and other modalities while more efficient, flexible, and scalable compared to ViT. thanks to Sujoy Paul, @gaganjain1582 , @vtresp , and @jainprateek_

[rajat koner @rajatkoner]

Excited to announce that our
@GoogleDeepMind
paper, "LookupViT: Compressing visual information to a limited number of tokens," has been accepted at
@eccvconf
three key takeaways...
https://t.co/Vu2V2GMVN7

[rajat koner @rajatkoner]

- Introducing high-resolution lookup tokens and smaller compressed tokens.
- Expensive self-attention and MLP operations are restricted on the compressed tokens.
- Updated information between compress and lookup tokens interact through a shared bidirectional cross-attention.

[rajat koner @rajatkoner]

@alaa_nouby @Apple interested...pinged u

[rajat koner @rajatkoner]

@silviaseverini3 @HinrichSchuetze @Bollegala @CisLmu Congrats @silviaseverini3 🎉

[rajat koner @rajatkoner]

@laoreja001 Is the internship till open?

[rajat koner @rajatkoner]

@laoreja001 Great opportunity, sent an email..hope m not very late to apply

[rajat koner @rajatkoner]

https://t.co/d0EvDrNheK

[rajat koner @rajatkoner]

If you are in Washington D.C. or attending @RealAAAI , check out our poster on 12th December named InstanceFormer, it is a simplified, tracker-free, single-stage video instance segmentation framework for long and occluded videos.@hannan_tanveer @sahandsharif

[rajat koner @rajatkoner]

For more information, please take a look at our paper or contact us: https://t.co/yKCRErfUW3 . See you at poster season. #AAAI23 #AAAI #AI #CV #Video_Understanding #Segemntation #tracking

[rajat koner @rajatkoner]

@stardazed0 great work👍

[rajat koner @rajatkoner]

#ECCV2022 presenting now “Relationfomer” it can generate insightful object-relation graph out of any image ..visit our poster in 36th counter..see u there ,paper: https://t.co/TSTJ04ppIU

[rajat koner @rajatkoner]

Delighted to be at @eccv2022 in Tel Aviv; if you are interested in a scene or object-relation graph from natural to medical image, please visit the poster "Relationformer" on the 27th at Hall B in the first half.@supro_s @sahandsharif @vtresp Full paper: https://t.co/TSTJ047Okk

[rajat koner @rajatkoner]

@sahandsharif @DeepMind Congrats @sahandsharif

[rajat koner @rajatkoner]

Preprint: https://t.co/TSTJ04pXys
Code: https://t.co/KM5Q0DvFhw

[rajat koner @rajatkoner]

Delighted to share our paper 'Relationformer: A Unified Framework for Image-to-Graph Generation.' got accepted in #ECCV2022. It is a single-stage "object- relation" generation method for scene graphs, 3D vessels, and road networks. Thanks to @supro_s ,@sahandsharif ,@vtresp

[rajat koner @rajatkoner]

@dutran great topics...applied..thanks for sharing

[Azad Hind @AzadHindAbad]

Whose image on new Rs.1000/- note would bestow the greatest honor upon the note?

[rajat koner @rajatkoner]

I can't go to India before March as given in RBI faq point 22 and 23 ! @SushmaSwaraj @arunjaitley @RBI @PMOIndia https://t.co/YebseWnKuc

[rajat koner @rajatkoner]

@SushmaSwaraj @RBI https://t.co/mKEZA4dYyy

[rajat koner @rajatkoner]

@SushmaSwaraj @RBI https://t.co/mKEZA4dYyy

[rajat koner @rajatkoner]

I am a student in Munich & hv few 500 notes.Where will I exchange them, as Forex counter refuses to take old notes @arunjaitley @PMOIndia