I will be attending ICML in Seoul next week.
We are presenting PRAGMA Revolut’s foundation model on one of the workshops (FMSD) on Saturday.
DM me if you’d like to meet to discuss financial foundation models or enjoy some great coffee in the city.
Noticed that Copilot feels less useful in .ipynb files compared to .py 🤔 🤔
Could it be the bloated .ipynb markup format? It's also why version-controlling these files is such a pain.
Surely there’s a way to make Copilot strip the clutter.
Attending #ICASSP2024 in Seoul? Don't miss our presentation on Synchformer (https://t.co/dcAiXR6mCM) this Tuesday at 13:10.
Also, let me know if you'd like to meet in Seoul next week!
📢 Thrilled to share that our paper on "Synchformer: Efficient Synchronization from Sparse Cues" got accepted at #ICASSP24!
🎉 Huge shoutout to the amazing team: @WeidiXie, Esa Rahtu, and Andrew Zisserman!
Code: https://t.co/z9cKvBBZ3D
arXiv: https://t.co/64I5N14V89
📢 Thrilled to share that our paper on "Synchformer: Efficient Synchronization from Sparse Cues" got accepted at #ICASSP24!
🎉 Huge shoutout to the amazing team: @WeidiXie, Esa Rahtu, and Andrew Zisserman!
Code: https://t.co/z9cKvBBZ3D
arXiv: https://t.co/64I5N14V89
Excited to share that I've joined @Oxford_VGG as a postdoc 🤖!
I'll be pursuing my passion for designing more advanced multi-modal models for video understanding 🔈🌆📽️🔬. Couldn't have asked for a better group to be a part of.
@y_m_asano Also, I would like to express my gratitude to my wonderful reviewers @andrewhowens and @RSerizel for their invaluable feedback and punctual submission of their reviews. Your support was crucial in making this project a success. Thank you!
Thrilled to share that I've successfully defended my PhD thesis, marking the end of an incredible journey. It was an honour to defend against such a brilliant and accomplished scholar @y_m_asano.
Also excited to announce that I'll be joining the VGG as a post-doc later this year
Just moved my web-based object detector app from a cloud instance to HuggingFace Spaces ❤️, and updated the interface to Gradio✨(prev HTML/CSS + JS mess 🦖)
Also, decided to check the upload trend over the years. Didn't expect it to gain this traction👀
https://t.co/rnUaKx9Rfo
Today marks one year since Putin's senseless war in Ukraine began. My heart goes out to the Ukrainian people who have suffered greatly. It's time to end this horrendous conflict! We must work towards a peaceful resolution and support Ukraine in its pursuit of sovereignty.
With the emergence of new chatbots like ChatGPT, text scraped from the web may become less useful for training LLMs.
At least, the 'fake/real' classifier will be a necessary part of the pre-processing pipeline.
Our paper was accepted at #BMVC2022 (Spotlight) 🎉
"Sparse in Space and Time: Audio-visual Synchronisation with Trainable Selectors"
Paper, code & colab: https://t.co/jScsGt9QX2
I couldn’t wish for better collaborators. Many thanks to @WeidiXie, Esa Rahtu, and Andrew Zisserman
I was taught that we have to remember WW2 and celebrate Victory Day every year to avoid such things from happening again. All these hypocritic parades and monuments on every second square in a town – "Remember" "Never joke around." Today the same people start the war.
Recently I had to take IELTS (high-stakes 🇬🇧 exam). It was my 2nd time and now computer-based (on-premise and w/ f2f speaking). Got 8/9. Confirming that the latest Cambridge IELTS books are the best to practice the structure.
This conceited smile on the registration photo tho…
Our paper "Taming Visually Guided Sound Generation" (🖼️👉🔉) was accepted as an oral presentation for the #BMVC2021 🎉.
Code, models, and a Colab demo: https://t.co/Yw2bvu1SOa
Project page, samples, and the gist: https://t.co/D6nx5lDtTh
Paper: https://t.co/vlLTeGZ42y
Leaving your for-loop outside of a function in Python makes it ~50% slower 😱. The issue has to do w/ the variable scope. Locals are stored as a fixed-sized array (fast indexing), globals as dicts (slower). Besides, 🐍 makes opcode prediction: squeezing several bytecodes into one
Sharing the development pipeline of the #YOLO Object Detector web app (👉 https://t.co/mTnlEncwrE)
I built 2y ago for fun👨💻
https://t.co/hQGzdu0gi4
Covers front-end (js, HTML), back-end (Flask, 🐍), how to rent a server, domain name, add it to a DNS, and secure the connection🔒
I'm keen on the opportunity to check out the NeurIPS paper reviews once I read one and formed an opinion of my own. These're useful as a fairly educated take on both pros and, most importantly, cons that are often only somewhat evident just from 1 pass. But mind the survival bias
Once I couldn't ssh to my machine. I forgot the tensorboard under tmux for 2 days and it ate up all RAM. It was running over the unchanged folder. Since then I prepend the cmd: `timeout 10h tensorboard…` (auto kill after 10h). Decided to share it because it just happened again