0asa (@0asa) header
0asa (@0asa) profile picture

Father, Innovation Engineer @EVSequipment, previously Data Scientist @Kensuio, E-Marketing Tools Manager @diagenode, cofounder @AntibodyAdviser.

Belgium
Joined May 2009
OpenAI, Anthropic, et al. scraped my website, my videos, my codebases, and all of the content I've published online. They used years of *my* work to train *their* models. Now, they claim this is "their data" and accuse others of "stealing" it. It's really hard for me to empathize with them about this. It's also hard to understand how they can claim moral superiority concerning the censuring of models. Both the GPT model series and Claude are totally nerfed and ideologically biased! Here I am, once more, rooting for free, open-science, and open-source models to be the last ones standing.
278
1,131
107
9,699
766,571
Thrilled to announce my first paper, "Dynamic NeRFs for Soccer Scenes", which was accepted at the MMSports 2023 workshop (@ ACM Multimedia)! It features a part of my master's thesis work, in which I explore the use of dynamic NeRFs for the task of reconstructing soccer replays.
Dynamic NeRFs for Soccer Scenes paper page: huggingface.co/papers/2309.0… The long-standing problem of novel view synthesis has many applications, notably in sports broadcasting. Photorealistic novel view synthesis of soccer actions, in particular, is of enormous interest to the broadcast industry. Yet only a few industrial solutions have been proposed, and even fewer that achieve near-broadcast quality of the synthetic replays. Except for their setup of multiple static cameras around the playfield, the best proprietary systems disclose close to no information about their inner workings. Leveraging multiple static cameras for such a task indeed presents a challenge rarely tackled in the literature, for a lack of public datasets: the reconstruction of a large-scale, mostly static environment, with small, fast-moving elements. Recently, the emergence of neural radiance fields has induced stunning progress in many novel view synthesis applications, leveraging deep learning principles to produce photorealistic results in the most challenging settings. In this work, we investigate the feasibility of basing a solution to the task on dynamic NeRFs, i.e., neural models purposed to reconstruct general dynamic content. We compose synthetic soccer environments and conduct multiple experiments using them, identifying key components that help reconstruct soccer scenes with dynamic NeRFs. We show that, although this approach cannot fully meet the quality requirements for the target application, it suggests promising avenues toward a cost-efficient, automatic solution. We also make our work dataset and code publicly available, with the goal to encourage further efforts from the research community on the task of novel view synthesis for dynamic soccer scenes
2
7
1
12
2,268
0asa retweeted
Dynamic NeRFs for Soccer Scenes paper page: huggingface.co/papers/2309.0… The long-standing problem of novel view synthesis has many applications, notably in sports broadcasting. Photorealistic novel view synthesis of soccer actions, in particular, is of enormous interest to the broadcast industry. Yet only a few industrial solutions have been proposed, and even fewer that achieve near-broadcast quality of the synthetic replays. Except for their setup of multiple static cameras around the playfield, the best proprietary systems disclose close to no information about their inner workings. Leveraging multiple static cameras for such a task indeed presents a challenge rarely tackled in the literature, for a lack of public datasets: the reconstruction of a large-scale, mostly static environment, with small, fast-moving elements. Recently, the emergence of neural radiance fields has induced stunning progress in many novel view synthesis applications, leveraging deep learning principles to produce photorealistic results in the most challenging settings. In this work, we investigate the feasibility of basing a solution to the task on dynamic NeRFs, i.e., neural models purposed to reconstruct general dynamic content. We compose synthetic soccer environments and conduct multiple experiments using them, identifying key components that help reconstruct soccer scenes with dynamic NeRFs. We show that, although this approach cannot fully meet the quality requirements for the target application, it suggests promising avenues toward a cost-efficient, automatic solution. We also make our work dataset and code publicly available, with the goal to encourage further efforts from the research community on the task of novel view synthesis for dynamic soccer scenes
2
31
4
167
29,729
Our AI-powered super slow-motion service just got a major upgrade with the ability to deploy on-prem, which means we've reduced the turnaround time to approximately five seconds! 👉 Read our press release: lnkd.in/e7AqvTqg #EVSforLive #nab #nab2023 #nab100 #ai
4
7
1,558
Raycast is blowing my mind. It's the 20% of macOS I didn't realize was missing. It's everything I wanted Alfred to be. It's clearly made by people obsessed with nailing the details. It's been 2 days and I can't remember how I used my computer before it.
44
80
4
870
269,639
Models such as Stable Diffusion are trained on copyrighted, trademarked, private, and sensitive images. Yet, our new paper shows that diffusion models memorize images from their training data and emit them at generation time. Paper: arxiv.org/abs/2301.13188 👇[1/9]
148
1,598
426
8,471
2,984,088
🤯 Full body tracking now possible using only WiFi signals A deep neural network maps the phase and amplitude of WiFi signals to UV coordinates within 24 human regions The model can estimate the dense pose of multiple subjects by utilizing WiFi signals as the only input 🧵
86
1,144
391
4,679
1,134,799
0asa retweeted
We spent $3,201,564.24 on cloud in 2022 at @37signals, mostly AWS. $907,837.83 on S3. $473,196.30 on RDS. $519,959.60 on OpenSearch. $123,852.30 on Elasticache. This is with long commits (S3 for 4 years!!), reserved instances, etc. Just obscene. Will publish full accounting soon.
161
307
116
2,259
734,959
0asa retweeted
Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition abs: arxiv.org/abs/2211.12368 project page: me.kiui.moe/radnerf/
17
233
27
1,143
Film Camera Trained a new model for 'realistic photos' - whatever that ever means. It works pretty well. Imagine where the quality will be in a year. #stablediffusion #portrait
53
196
62
1,491
Turned my daughter’s Moana doll into a 3D animated model. 1) @LumaLabsAI -> NeRF 2) @Blender -> mesh cleanup 3) @Mixamo -> rigging and animation 4) AR Quick Look Now dreaming up ideas for a Toy Story experience with all her favorite toys. The future of entertainment is bright!
14
76
10
515
Things could be better zpr.io/EeP6qFyVHbnU
17
175
119
691
Replying to @sama
7,987
21,675
3,126
245,844
FTX operated like a bank. That’s the problem. Crypto has nothing to do with this entire shitstorm. Wait till they learn that all banks operate with fractional reserves.
104
205
10
1,197
I think this 19th century bear hunting armor needs to be refreshed for crypto
95
130
13
1,655
"Language is an imperfect, incomplete, and low-bandwidth serialization protocol for the internal data structures we call thoughts." (Yann LeCun)
Wondering how to use yesterday's release? Here are some tips and tricks for making the most of Frame Interpolation ⚡️
1
24
4
207
"Bankers are looting the world. You're not in the middle of a recession; you're in the middle of a robbery." - Frankie Boyle
210
2,920
143
10,657
Spent the past days in Liège at @EVSEquipment to get Training on the LSM and LSM-VIA. After only knowing the basics of clip creation and shuttling back and forth it was great to learn the Ins and Outs of the system! If you're looking for a LSM Operator feel free to reach out ❤️
1
6