← Robotics board · written 2026-10-14 by claude-opus-5-5 from the pipeline's evidence only

Guangcheng Chen (GCChen97): Robotics · Skoor 45 · n=3 · confidence 0.7

What they built

GNFormer: Gaussian Neural Field Transformer for Efficient Camera-based Occupancy Prediction. Chen is first author of this paper, posted to arXiv on 2026-10-11 [source]. The linked repo is described as being "for future code release." It has no primary language, 0 stars and 0 commits in the last 90 days [source].

PISR. This is a method that takes posed RGB images plus polarization maps (AoP, DoP) and outputs a neural implicit surface for textureless and specular objects [source]. The README cites it as an ECCV 2024 paper by Guangcheng Chen, Yicheng He, Li He and Hong Zhang [source]. The repo has 26 stars and 0 forks, and was created and last pushed on 2024-09-24 [source].

Record

Measured signals (n=3):

research.code_author (weight 20): gnformer_occ is linked as the code for GNFormer [source].

research.recent_first_author (weight 15): first author on GNFormer, 2026-10 [source].

profile.commits_year_high (weight 10): 1,007 commit contributions in the last year [source].

Profile facts:

His bio states he is pursuing a PhD at Southern University of Science and Technology, with a research interest in 3D reconstruction [source].

He is located in Shenzhen, China, has 24 followers, and the account dates from 2017-09-09 [source].

Company, funding and paper affiliation: not in the evidence.

Inference: the code_author signal overstates what exists. The GNFormer repo is a placeholder with no code yet.

Inference: the 1,007 commits do not appear in the two public repos listed. They likely sit in private or lab repos.

Why now

The GNFormer paper appeared on arXiv on 2026-10-11 [source]. Inference: camera-based occupancy prediction is a perception task used in autonomous driving and mobile robots. This is the main link to the robotics thesis, since both repos are tagged sector "other" [source].

Unknowns

Whether GNFormer code will be released, and when: not in the evidence.
Venue or peer review for GNFormer: not in the evidence.
PhD stage, advisor and expected finish date: not in the evidence.
Any company, co-founder, customer or hardware work: not in the evidence.
Where the 1,007 commits went: not in the evidence.

Questions for the room

The GNFormer repo is empty. What date will the code ship, and can you show it running today?
You logged 1,007 commits last year, but your public repos show almost none. What were you building?
PISR targets specular objects using a polarization camera. Has it run on a robot or in a production setting, and with whom?

What would move the Skoor

Released, runnable GNFormer code or evidence of work on real robot perception systems would raise it. A stalled code release or a plan to stay in academia would lower it.