Lei Kang

PostDoc@CVC_UAB . Interested in #DiffusionModels #DocumentAnalysis #HumanDocumentInteraction . Views my own.

[Lei Kang @lei_tro]

“An Effective Data Augmentation Method by Asking Questions About Scene Text Images” by CVC intern Xu Yao!
📍 Poster PA-17.4
🗓️ May 7, 2:00–4:00 PM
We show how question-driven augmentation can improve scene text understanding. Stop by and chat! #ICASSP 
https://t.co/KDERMJSYDY

[CVC_UAB @CVC_UAB]

🚀 Workshop Announcement – ICPR 2026

#CVC, along with @UABBarcelona, Università degli Studi di Cagliari (@univca) , @UniGenova, and AllRead MLT, is excited to organize TrustDOC: Trustworthy Document Understanding – Privacy, Unlearning, Robustness, and Explainability.

📅 Friday, August 21, 2026 | 8:30 AM – 12:30 PM
📍 International Convention Center, Lyon, France

This workshop, endorsed by International Association for Pattern Recognition (@ExcoIapr) and TC11 and supported by ELSA – European Lighthouse on Secure and Safe AI(@elsa_lighthouse), will explore the latest advances in trustworthy document understanding.

📌 Important Dates:
Paper Submission: 16/05/2026 CET
Notification to Authors: 11/06/2026 CET
Camera-ready Deadline: 18/06/2026 CET
Workshop: 21/08/2026 CET

📄 Submit your papers via CMT: https://t.co/h6V0AWvbZG
🌐 More info: https://t.co/JFp6rXCLjg

[Lei Kang @lei_tro]

Wow @HochreiterSepp shared our arXiv paper! Super grateful and exited! xLSTM has worked great for us in our tasks and we recommend others give it a shot! Cheers🍺🍺🍺

[Sepp Hochreiter @HochreiterSepp]

xLSTM for Multi-label ECG Classification: https://t.co/rGUahBRC4W

"This approach significantly improves ECG classification accuracy, thereby advancing clinical diagnostics and patient care."
Cool.

[Sophia Yang, Ph.D. @sophiamyang]

Announcing @MistralAI OCR -  the world’s best document understanding API.

🔍 State-of-the-art understanding of complex documents
🌍 Natively multilingual and multimodal
⚡ Fastest in its category
📄 Doc-as-prompt, structured output
🔒 Available for on-prem deployment

[ELLISBarcelona @ELLISBarcelona]

🗓️ The ELLIS Unit Barcelona is co-organising the Deep Learning Barcelona Symposium (@dlbcnai).

Our co-director, Dimosthenis Karatzas (@dkaratzas), has opened the first oral session of the day alongside Marco Beroni from @icreacommunity.

[Zhiyu Zoey Chen @ZhiyuChen4]

I'm shocked to see racism happening in academia again, at the best AI conference @NeurIPSConf. Targeting specific ethnic groups to describe misconduct is inappropriate and unacceptable. @NeurIPSConf must take a stand. We call on Rosalind Picard @MIT @medialab to retract and apologize for her statement.

[Lei Kang @lei_tro]

@fofrAI Like a real hand drawing💯

[Lei Kang @lei_tro]

ArXiv: https://t.co/30n7w2b9D6
GitHub: https://t.co/RUi2aSJDV1
Welcome to open an issue on our GitHub repo or reach out to us via email for further discussion. Cheers!🍻🍻🍻

[Lei Kang @lei_tro]

Hi #ecai2024 ers! 🐙We’ll present our work, GRIF-DM: Generation of Rich Impression Fonts Using Diffusion Models, tomorrow Wednesday from 11:15-11:30 in room R11 Budapest. Work done with Fei Yang, @Alvaro_kai93 @SouibguiAli , Lluis Gomez, @AliciaFornes @ErnestValveny @dkaratzas .

[Maura Pintor @maurapintor]

AISec 2024 update:
We're delighted to share our complete workshop schedule, featuring 18 accepted papers, engaging poster sessions, and three distinguished speakers in AI security.
Visit https://t.co/eDczCRRGJf to explore the program!

[Lei Kang @lei_tro]

The time has been modified to 11:50 - 12:10 🤣🤣🤣

[Lei Kang @lei_tro]

Hi #icdar2024 ers! We’ll present our work on Machine Unlearning for Document Classification from 12:10-12:30 on Sept 2 in oral session II: Layout Analysis and Doc Class. Our work is done with @SouibguiAli , Fei Yang, Lluis Gomez, @ErnestValveny , and @dkaratzas . See you then!

[Lei Kang @lei_tro]

arXiv: https://t.co/hZcFCcfviJ
GitHub: https://t.co/g2TkmiKT0j

[Lei Kang @lei_tro]

Looks cool! To a VQA task, it can generate both answers and bounding boxes. I can see its potential to be served as an evidence finder🤟

[岡本大和 / Yamato.OKAMOTO / RoadRoller @RoadRoller_DESU]

CREPE's ability to output structured data widens its applicability. It can handle tasks like VQA and simultaneous analysis of multiple documents. Beyond document analysis, for instance in scene images, it can perform object detection while simultaneously recognizing text.

[Lei Kang @lei_tro]

@d_hintersdorf @ecai2024 Well done! Nice work!!

[Lei Kang @lei_tro]

It’s a really good way to understand how a transformer learns the inherent functioning of a “calculator” w/o external mathematical knowledge. But when looking at the comments, qué pasó jaja?

[Sean McLeish @SeanMcleish]

Introducing 🧮Abacus Embeddings, a simple tweak to positional embeddings that enables LLMs to do addition, multiplication, sorting, and more. Our Abacus Embeddings trained only on 20-digit addition generalise near perfectly to 100+ digits. 1/n

[CVC_UAB @CVC_UAB]

📢📢 𝟒 𝐝𝐚𝐲𝐬 𝐥𝐞𝐟𝐭 to participate in the #NeurIPS2023 Privacy Preserving Document Understanding Challenge, #PFLDocVQA🚀

⚠ Submit your data before 𝐅𝐫𝐢𝐝𝐚𝐲, 𝐎𝐜𝐭𝐨𝐛𝐞𝐫 𝟐𝟕!

#PrivacyPreservation #AIChallenge #DocumentAnalysis #DocumentUnderstanding

[CVC_UAB @CVC_UAB]

📣 Exciting News! The #NeurIPS2023 Privacy Preserving Document Understanding Challenge is here!🚀

👉 The competition focuses on potential privacy leakage problems when dealing with sensitive information using state-of-the-art #DocumentAnalysis techniques and LLMs.

[Lei Kang @lei_tro]

Big thanks to my supervisors @MRossinyol @AliciaFornes @mauricio_ville , I can never go this far without you🤗🤗🤗 Also sincere gratitude to my thesis committee for the time and effort! Salud!!!🍻🍻🍻

[CVC_UAB @CVC_UAB]

🎓We have a new Doctor at CVC!!! @lei_tro defended yesterday his thesis “Robust Handwritten Text Recognition in Scarce Labeling Scenarios: Disentanglement, Adaptation and Generation”. Congratulations Dr. Lei Kang! All the best for the future! 👏🥳🎉
▶️https://t.co/rllS9hBGbd

[Lei Kang @lei_tro]

It's my first time to step into the office since March, quite a long time! And super happy to fetch the Best Paper Award diploma of #ICFHR2020 ! A big thanks to the document analysis community and salud to my colleagues🍺🍺🍺