Ishan Dave 🇮🇳

Sr. Applied Scientist @Adobe, Ph.D. in Computer Vision (Videos) from UCF. Ex-Apple, Adobe Research, IISc, NIT

[Ishan Dave 🇮🇳 @_ishucf_]

Had one of my best conference experiences at #ECCV2024, with 4 major presentations: 1 oral, 2 posters, and 1 doctoral consortium. It was an honor to interact with incredible researchers during my sessions: Dr.Efros,@dimadamen @doughty_hazel @arnoldsmeulders @JCollomosse @cgmsnoek

[Ishan Dave 🇮🇳 @_ishucf_]

@webalorn Great work! Would appreciate a video results link if possible.

[Ishan Dave 🇮🇳 @_ishucf_]

@TzDev_ Upscale

[Ishan Dave 🇮🇳 @_ishucf_]

Oral Talk
📍 Exhibit Hall III
🕒 Tue, Oct 21 — 9:00–9:15 a.m. HST (3:00–3:15 p.m. EDT)

Poster Session
📍 Exhibit Hall I
🕕 Tue, Oct 21 — 11:45 a.m.–1:45 p.m. HST (5:45–7:45 p.m. EDT)

📄 arXiv: https://t.co/TPKF3c5iZi
🌐 Project Page: https://t.co/V2BoEHBCr0

[Ishan Dave 🇮🇳 @_ishucf_]

🎉 Our work GT-Loc has been accepted as an Oral Presentation (🌟 Top 64 of 11,239 submissions) at #ICCV2025 ! It introduces a unified framework for geo-localization & time-of-capture prediction, learning a joint embedding to retrieve when & where an image was taken.🗺️📸@ucfmshah

[Ishan Dave 🇮🇳 @_ishucf_]

@CSProfKGD and Prof. Efros vibing with our Alignable Video Retrieval poster- because who doesn’t love perfect alignment? 😄#ECCV2024
Don't miss project page: https://t.co/rN1EZW2xkA

[Kosta Derpanis @CSProfKGD]

Thanks for the overview @_ishucf_! #ECCV2024

[Ishan Dave 🇮🇳 @_ishucf_]

Here’s an overview of the works from my Ph.D. thesis, titled 'Enhancing Label Efficiency and Privacy in Video Understanding'. More details can be found on my personal page: https://t.co/S2i3rroWXA #ECCV2024

[Ishan Dave 🇮🇳 @_ishucf_]

Had an amazing experience presenting my Ph.D. thesis work (advisor @ucfmshah) at the Doctoral Consortium at #ECCV2024. Big thanks to @cgmsnoek for his mentorship throughout the program.

[Ishan Dave 🇮🇳 @_ishucf_]

Grateful to @CVPR for this honor! Thrilled to be recognized as an outstanding reviewer! 😇
@UCFCRCV @ucfmshah

[#CVPR2027 @CVPR]

HUGE shoutout to our #CVPR2024 Outstanding Reviewers 🫡

[Ishan Dave 🇮🇳 @_ishucf_]

Project page: https://t.co/4YoPMwX0mc
ArXiv: https://t.co/GdsgzJ7QHP
YouTube presentation: https://t.co/0bePbhcRLc

[Ishan Dave 🇮🇳 @_ishucf_]

Our AAAI paper, "No More Shortcuts: Realizing the Potential of Temporal Self-Supervision", is all set for presentation today (6:15 pm, poster#61) in the main track. This research was part of my internship at @AdobeResearch and with my advisor @ucfmshah. @RealAAAI @UCFCRCV #AAAI24

[Ishan Dave 🇮🇳 @_ishucf_]

Our key contributions include identifying and removing local-patch-based shortcuts in temporal pretext tasks, aiding in learning better temporal correspondence between frame features. We also propose more challenging frame-level pretext tasks. The image shows overall framework:

[Ishan Dave 🇮🇳 @_ishucf_]

Grateful for 2023: Proud to mentor juniors on their first publications at ICCV, ICRA, & IROS, and to celebrate my own work published at CVPR and accepted for AAAI'24. Immense gratitude to my advisor @ucfmshah and lab mates. Eagerly anticipating an even more fruitful 2024!

[Ishan Dave 🇮🇳 @_ishucf_]

@dimadamen @doughty_hazel @CVPR @cgmsnoek Proud to see our TCLR highlighted in your keynote!😇
https://t.co/SMj5vDEK2d
@rohitgUCF @MamshadR @ucfmshah

[Ishan Dave 🇮🇳 @_ishucf_]

Due to VISA delays, I'll be attending #CVPR2023 virtually! My labmate @joefioresi5 will be representing us in person at our Poster Session tomorrow: TUE-AM-224. Looking forward to insightful discussions online and offline.

[Ishan Dave 🇮🇳 @_ishucf_]

🎉 Exciting to announce #CVPR2023 work (w/ @MamshadR @ucfmshah @UCFCRCV): TimeBalance, a new approach for Semi-Supervised Learning in Action Recognition! Access details here:📄https://t.co/13daCY54dR 🎥https://t.co/m1EHWwSw7I 🌐 https://t.co/q0QgjkHaRY #ComputerVision #CVPR

[Ishan Dave 🇮🇳 @_ishucf_]

Our method shows top-notch performance, achieving state-of-the-art on action recognition benchmarks like UCF101, HMDB51, and Kinetics400 beating multistream approaches like Video+Optical Flow/Gradients with just RGB stream #ActionRecognition