Lab1055 @ IITH

Learning robust and reliable systems in evolving environments. The machine learning and computer vision group at @IITHyderabad!

[Lab1055 @ IITH @lab1055]

07th June
𝗩𝗲𝗻𝘂𝗲: Exhibit F
𝗣𝗼𝘀𝘁𝗲𝗿 𝗕𝗼𝗮𝗿𝗱: 75

[Lab1055 @ IITH @lab1055]

𝗙𝗔𝗟𝗖𝗢𝗡-𝗦𝗙𝗢𝗗: 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗠𝗼𝗱𝗲𝗹 𝗣𝗿𝗶𝗼𝗿𝘀 𝗘𝗻𝗵𝗮𝗻𝗰𝗲 𝗢𝗯𝗷𝗲𝗰𝘁 𝗙𝗼𝗰𝘂𝘀 𝗶𝗻 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗦𝗽𝗮𝗰𝗲 𝗳𝗼𝗿 𝗦𝗼𝘂𝗿𝗰𝗲-𝗙𝗿𝗲𝗲 𝗢𝗯𝗷𝗲𝗰𝘁 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻

https://t.co/wPlLR8jF3S

[Lab1055 @ IITH @lab1055]

Work by: @sairamvcr, Rishabh Lalla, @aveen_Dayal, Tejal Kulkarni, Anuj Lalla, @nbvineeth, Muhammad Haris Khan

[Lab1055 @ IITH @lab1055]

FALCON-SFOD addresses this by using foundation-model priors to restore foreground-focused feature representations, leading to more robust source-free adaptation.

[Lab1055 @ IITH @lab1055]

📢[CVPR '26] What really breaks under domain shift in source-free object detection? We show, for the first time, that detectors often lose object focus and start activating on background clutter in the feature space.

[CSE IITH @cse_iith]

𝗣𝗮𝗽𝗲𝗿 𝗔𝗰𝗰𝗲𝗽𝘁𝗮𝗻𝗰𝗲 𝗔𝗻𝗻𝗼𝘂𝗻𝗰𝗲𝗺𝗲𝗻𝘁 🎉

Paper titled "Transitioning Heads Conundrum: The Hidden Bottleneck in Long-Tailed Class-Incremental Learning" has been accepted at TMLR 2026 (Transactions on Machine Learning Research).

Authors: Rahul Vigneswaran, Hari Chandana Kuchibhotla, Vineeth N Balasubramanian

👏 Congratulations to all the authors!

🔍 Key Highlight:
This work introduces DEREK (DEcoupling Representations for Early Knowledge Distillation), a method addressing a previously overlooked challenge in Long-Tailed Class-Incremental Learning (LTCIL): the Transitioning Heads Conundrum.

In LTCIL, head classes that are well-represented in earlier tasks become tail classes in subsequent tasks due to memory constraints, leading to accelerated catastrophic forgetting. DEREK mitigates this by decoupling head and tail learning via specialized expert networks and applying Early Knowledge Distillation before data constraints take effect, preserving rich representations.

Across 2 LTCIL benchmarks, 12 experimental settings, and 24 baselines, DEREK consistently establishes new state-of-the-art performance.

#MachineLearning #ContinualLearning #LongTailedLearning #KnowledgeDistillation #TMLR2026 #IITHyderabad

[Lab1055 @ IITH @lab1055]

Hearty congratulations to all authors: @AgrawalSus35924, Krishn V Kher, Saksham Mittal, @swarniminloop, @nbvineeth. The authors will be presenting it at NeurIPS 2025 this December.

[Lab1055 @ IITH @lab1055]

Excited to share that our paper “Memory-Integrated Reconfigurable Adapters (MIRA)” has been accepted at NeurIPS 2025 🚀

https://t.co/ZpLbi7HfYr

[Lab1055 @ IITH @lab1055]

MIRA combines Hopfield-style associative memory with lightweight adapters, enabling a single model to tackle domain generalization, continual learning, and class-incremental learning, achieving state-of-the-art performance across multiple benchmarks.

[Lab1055 @ IITH @lab1055]

Hot off the press from @somethingmyname, Sai Srinivas, Bamdev Mishra, @nbvineeth

[Rahul Vigneswaran @somethingmyname]

🚨 My paper got accepted into TMLR 25! 🚨

Are you a victim of your loan being rejected by an algorithm? Did the countermeasures provided by your bank seem unreasonable?

Ask your bank to adopt HARE! 🐇✨

Paper: https://t.co/PttJI6FHU7

1/5
@TmlrOrg @TmlrPub @cse_iith

[Lab1055 @ IITH @lab1055]

This is a joint work by @AgrawalSus35924, @VemuriDeepika, Siddarth Chakaravarthy, & @nbvineeth.

[Lab1055 @ IITH @lab1055]

🚀 Poster at @RealAAAI! Walking the Web of Concept-Class Relationships

🔍 Explores how concept-class relationships degrade in incremental learning settings

📍 AAAI, Exhibit Hall E, March 1st, 12:30 EST. Susmit will be at the poster session!
📖 Read here: https://t.co/tRpQ6BxyS8

[Lab1055 @ IITH @lab1055]

📰Improving Unsupervised Domain Adaptation: A
Pseudo-Candidate Set Approach | A work carried out by: @aveen_Dayal, @Rishabh_Lalla, Linga Reddy Cenkeramaddi, C. Krishna Mohan, Abhinav Kumar, @nbvineeth

https://t.co/4Xn1osHF9t

[Lab1055 @ IITH @lab1055]

🎉 In our latest work @eccvconf we introduce the concept of a pseudo-candidate set-- a first in UDA-- to improve UDA methods across the board! Catch up with @aveen_Dayal at Poster Session 3, Oct 2, 10:30-12:30 (Board ID 6). Check out the video: https://t.co/41PJw5ytHR #ECCV2024

[Lab1055 @ IITH @lab1055]

From the #XAI4CV workshop at #CVPR 2024!

[Sukrut Rao @sukrutrao]

Our final invited talk is by Prof. Vineeth N Balasubramanian (@lab1055), who talks about "Moving beyond an Afterthought: Towards Learning via Explanations".

[Lab1055 @ IITH @lab1055]

Shout-out for organizing workshops at ACCV 2024 at Vietnam. Proposal deadline: May 31st. It is a good opportunity to run your workshop, promote your research areas🚀, connect with others👩‍🎓, and travel 🛬to a beautiful country like Vietnam 🇻🇳 @ACCVConf

https://t.co/MIuW5FAxIB

[ACCV @ACCVConf]

📢 Calling all community developers and visionaries! @ACCVConf workshop proposals are open.
Don't miss the May 31 deadline!

Excited to see what you bring to the table in Hanoi, Vietnam. 🌟🇻🇳 #ACCV

https://t.co/HVO3li6A9J

[Lab1055 @ IITH @lab1055]

🚁 C2FDrone (ICRA 24 Oral), leverages objectness information embedded in the image representations to localise drones in drone-to-drone detection task.

Curious about how we do it?
Catch up with our PhD student, @RVCSR29 at @ieee_ras_icra for a chat!

https://t.co/LrVXPPDr14

[Lab1055 @ IITH @lab1055]

Like to know about a generalized method for causal treatment effect estimation? Read on about our method "NESTER", presented at AAAI 2024 @RealAAAI, that proposes a neurosymbolic approach that brings together existing ideas in a single framework at:
https://t.co/eTknCGDwYy

[Lab1055 @ IITH @lab1055]

If you're at @RealAAAI, and curious about applying GANs for ante-hoc explainability, catch @VemuriDeepika from our group at her poster session today!