Abdallah DIB

I am happy to share with you, FaceNext, an open source, lightweight @PyTorch library for high fidelity 3D face reconstruction from single/multiple RGB image(s). 1/ github.com/abdallahdib/Ne… #machinelearning #deeplearning #learning #DataScience

We will be presenting our work on creating 3D relightable avatars from portrait images at #cvpr2024. If you are attending, and wants to learn more or simply have a chat, i will be happy to meet u at poster session on wednesday June 19 10AM-12AM! See you @CVPR #cvpr

Our paper on 3D face reconstruction was accepted to #CVPR2024. We are releasing a new dataset 'FFHQ-UV-Intrinsics' available for download here: github.com/ubisoft/ubisof… This dataset contain intrinsic face attributes for 10k subjects. More details: ubisoft-laforge.github.io/character/mosa…

@pfau Indeed It looks simple as u describe it, but it opens new perspective for multi tasking which by itself is a great step forward. I think this is what makes this work amazing

I genuinely don't understand why people seem so excited by the GATO paper. They took a bunch of independently trained agents, and then amortized all of their policies into a single network? That doesn't seem in any way surprising given what we know about Transformers.

@_AlecJacobson i had similar weird things many times and its always funny to see. But for optim/learning, scaling Basis by the corresponding eigen values is useful so to make all coeffs lie in the same range which makes optim much stable but needs careful regularization to avoid such scenario

Accidentally scaled my principal component vectors, wasn't expecting this and really made me laugh

@BolkartTimo @DanecekRadek @Michael_J_Black very nice work Timo! I enjoyed looking at ur results, emotion reconstruction looks appealing. this solves a big limitation of existing face reconstruction methods...

#CVPR2022: Given a single image, EMOCA reconstructs a 3D face with facial expression detail that convey the emotional state of the input. 1/6 Code & model: github.com/radekd91/emoca PDF: ps.is.mpg.de/uploads_file/a… Video: youtu.be/zjMLB2-dVGw With @DanecekRadek @Michael_J_Black

@pacile_art @PyTorch Working on it :)

@abdallah_dib @PyTorch Can you set a google colab for this? It would be so usefull!

I am happy to share with you, FaceNext, an open source, lightweight @PyTorch library for high fidelity 3D face reconstruction from single/multiple RGB image(s). 1/ github.com/abdallahdib/Ne… #machinelearning #deeplearning #learning #DataScience

Trying NextFace on challenging image, still get meaningful reconstruction! github.com/abdallahdib/Ne… #MachineLearning #DeepLearning #AI #computervision #3D #rendering #PyTorch #Python

@PyTorch Features: - Reconstructs face at high fidelity from single or multiple RGB images - Estimates face geometry - Estimates detailed face reflectance (diffuse, specular and roughness) - Estimates scene light with spherical harmonics - Runs on both cpu and cuda-enabled gpu 2/

The Fourier Transform explained in one sentence [source and read more, by Stuart Riffle: buff.ly/3ILfsdD]

Instant Neural Graphics Primitives with a Multiresolution Hash Encoding paper: nvlabs.github.io/instant-ngp/as… project page: nvlabs.github.io/instant-ngp/ github: github.com/NVlabs/instant…

Don’t miss the multiple open PhD and Postdoc positions in my group #CVI2 @SnT_uni_lu @uni_lu in partnership with industrial partners @Artec3DScanners @LMO_Space @POST_Luxembourg and more coming soon…

@AlexGDimakis good point. I think some conf already do this (siggraph).

While waiting for #CVPR2022 CMT to get up again, I would like to propose a simple cryptographic solution to the big data submission problem: We only upload a SHA256 hash of our to-be-submitted pdf and then upload the committed pdf any time next week.

#CVPR2022 this not the first time that this happened with @MicrosoftCMT . Minor corrections are always important for a smooth reading.

Light... Is... Everything

In anticipation of the Intl. Conf. on Computer Vision (#ICCV2021) this week, I rounded up all papers that use Neural Radiance Fields (NeRFs) represented in the main #ICCV2021 conference here (1/N): dellaert.github.io/NeRF21

We will present our work on high fidelity 3D face reconstruction using self-supervised learning and ray tracing at #ICCV2021 during session 10 Wednesday 13Oct: 5–6PM (EDT time) Friday 15Oct: 10–11AM (EDT time) arXiv: arxiv.org/pdf/2103.15432… video: youtube.com/watch?v=VVr_bb…

Uh huh.

China overtakes US in #AI research asia.nikkei.com/Spotlight/Data…

Our work on monocular 3D face reconstruction has been accepted at #ICCV2021 We achieve high fidelity face reconstruction using self-supervised learning and differentiable ray tracing Video: m.youtube.com/watch?v=VVr_bb… Paper: arxiv.org/abs/2103.15432