Finally the day has arrived, I met my hero Ross Girshick (META) at #cvpr2023
Beginning from Faster rcnn, I have been following his research. Also got to know about his early days of object detection research, deformable part models (pre deep learning era) 1/n @CVPR
Second, I presented our work, Towards Calibrating Prompt Tuning of Vision-Language Models, which focuses on calibration for prompt-tuned vision-language models.
Link: https://t.co/dGBbz5e0Yw
n/n
CVPR 2026 wrap-up from Denver, Colorado -- an inspiring week of research discussions and new ideas. I had the opportunity to present two works this year.
1/n
First, I gave an oral talk at the MONTI workshop for our hashtag#BestPaperOral. ThinkGeo is our benchmark effort toward tool-augmented Earth Observation reasoning.
Link: https://t.co/hOxgQiVOg3
2/n
π HPC-AI is heading to #CVPR2026! Find us at Booth 237 from June 3β7! Whether you're building AI Agents or hunting for top-tier compute, weβve got you covered.
π The Ultimate CVPR Swag & Perks:
1οΈβ£ Limited Swag: Grab custom caps, canvas bags π§’ ποΈ
2οΈβ£ Top-up Reward: π°$15 = GPU Hour + Model APIs + Summer mini-fans / Bluetooth speaker (Only 20 set)π₯
3οΈβ£ Accepted Author Special: Top up $40 Get $20 voucher π°
4οΈβ£ Survey Gift: Just do a quick survey, choose your refrigerator magnets π§²
Stop by, talk tech, and claim your rewards! π€
π Booth 237 | June 3β7 | Colorado Convention Center
#CVPR2026 #HPCAI #AI #MachineLearning #GPU #ModelAPIs
Highlights:
436 diverse, human-curated queries across 7 domains
14 robust tools
Rigorous step-by-step and end-to-end evaluation of top LLMs (GPT-4o, Qwen, LLaMA-3)
Fully open-source
A big shoutout to the incredible team behind this work!
n/n
πWhy ThinkGeo?
Most existing agentic benchmarks focus on synthetic or generic scenarios. Remote sensing demands spatial and temporal reasoning over real-world satellite and aerial imagery. ThinkGeo bridges this gap by offering first agentic benchmark grounded in RS imagery. 2/n
Behind every great conference is a team of dedicated reviewers. Congratulations to this yearβs #CVPR2025 Outstanding Reviewers!
https://t.co/z8w4YJKTep
Exciting update about our #CVPR2025 paper: "O-TPT: Orthogonality Constraints for Calibrating Test-time Prompt Tuning in Vision-Language Models" has been selected as a highlight! π
Congratulations to all co-authors
Paper: https://t.co/wIrX06FLUn
Code: https://t.co/dWwnk1d7eT
1. O-TPT: Orthogonality Constraints for Calibrating Test-time Prompt Tuning in Vision-Language Models
Authors: Ashshak Sharifdeen, Muhammad Akhtar Munir, Sanoojan Baliah, Salman Khan, Muhammad Haris Khan
2/n
When I was a grad student, sometimes I felt isolated because there wasn't much collaboration among students and opportunities to work with different mentors.
So, three students from my lab will be visiting Berkeley, Stanford, and Cornell next Spring! We will miss having them around, but I am sure that the experiences will help them grow as better researchers!
Also, thanks to our friends for hosting them!
Announcing GitHub Copilot Free!
A new free tier for GitHub Copilot, available for everyone today in @code
No trial. No subscription. No credit card required.
Learn more in our blog: https://t.co/k3S9M9xefa
Applications for young researchers (undergrad, PhD or postdoc students) to attend the 2025 Heidelberg Laureate Forum
@HLForum
are now open. I love interacting with the young researchers there, and I'm hoping to be able to go this year, after not being able attend the last few years.
Links are in my first reply (trying an experiment since it seems posts with outgoing links are being throttled on this platform). β¬οΈ
Here is an updated version of my chart showing the progress of AI-weather models over time. For details, including raw data and links to sources, check out this spreadsheet (feedback welcome π): https://t.co/i0Sg61gox7
Our study adapts neural weather forecasting models for the MENA region using Low-Rank Adaptation (LoRA), leading to faster convergence and improved accuracy in predicting weather variables compared to a global model.
n/n
Pleased to share, our work "Efficient Localized Adaptation of Neural Weather Forecasting: A Case Study in the MENA Region" has been accepted at #NeurIPS2024 Workshop on Tackling Climate Change with Machine Learning
More: https://t.co/FM6z4PWf6y
w/ Fahad Khan @KhanSalmanH
1/n