Hey researcher friends! How are you discovering interesting papers outside of your network these days? @karpathy’s arxiv-sanity-lite.com? @labmlai’s overview papers.labml.ai? Something else? :)
Christoph Lassner
70 posts
Christoph Lassner
@chlassner
Co-founder of World Labs; spatial intelligence for 3D and 4D perception, rendering and generation.
- Couldn't be more excited to share World Labs today, our journey towards Spatial Intelligence! We are making the world your oyster with @drfeifei, @BenMildenhall, @jcjohnss and a world class team - more to come!Hello, world! We are World Labs, a spatial intelligence company building Large World Models (LWMs) to perceive, generate, and interact with the 3D world. Read more: worldlabs.ai/about
- Mark showed some really exciting tech today at #metaconnect2022 ! With my team @BozicAljaz, @denglide, @LukeDoukakis we created the object reconstruction and rendering using neural radiance fields youtu.be/hvfV-iGwYX8?t=… and can share a few more details.
- youtube.com/watch?v=gUARgo… Demoing our real-time #NeRF and object reconstruction to @jetscott from @CNET - fully interactive in #VR! Thanks for the kind words! Co-presented with the excellent inverse-rendering work by @flycooler and team.
- Replying to @chlassnerThis technology expands the range of objects that can be scanned very far and is very easy to use, yet produces incredibly detailed results that were impossible to create before---so you can take with you, share or present objects in VR with true-to-life fidelity.
- Replying to @wenzeljakob @seb_spe and @DelioViciniI just searched a lot for different options to write high performance (auto-)differentiable CUDA code and did not find frameworks that satisfied the constraints I had in mind. Will check out Dr. Jit!
- Replying to @chlassnerWhy is this exciting? Photogrammetry has been used in the past to create assets for CG/games, but could not be used for certain object classes (for example, objects with fur such as the teddy bear or semi-transparent objects).
- Does anyone have experience with using Rust-GPU (buff.ly/3Vsrdhr) for larger or performance-critical projects? How does the performance compare to CUDA? Any other interesting initiatives for language embedded and platform independent GPGPU programming?
- Getting ready to head out to NeurIPS tomorrow. If you’re around and would like to chat in person, let me know! See you in Vancouver!
- RT @[email protected] Dear network: do you know who is working on the macOS dynamic linker (/usr/bin/ld) at #Apple? Not to be overly dramatic, but it has started warning about an issue that (if true) will soon break every compiled Python extension (NumPy, PyTorch, (1/2)
- Replying to @chlassnerThese are results of a fully automatic pipeline - what you see is what you get purely from a phone video with no depth information, markers, etc., and there has been no artist editing, refinement or cleanup.
- Replying to @chlassnerWe manage to reconstruct the semi-transparent parts of the objects from the phone video directly and, thanks to volumetric rendering, they are composited correctly with the environment and other objects. The scientific fine-print:
