Yana · Computer vision
Gyanateet
I completed the MSc Advanced Computer Science (Artificial Intelligence) at the University of Leeds. I am a research technician at Leeds. From March to November 2025 I worked with AIMS (AI in Medicine and Surgery) on DINOv2 models for ESD workflow recognition.
ORCIDGoogle ScholarXISBI 2026W&BLinkedIn
Projects
- GOT-JEPA surgical tool tracking2026
Public code for GOT-JEPA multi-object tracking of surgical tools on CholecTrack20, separate from the phase-recognition paper.
- MVA Rare Disease Hackathon 20262026
A genome-to-genotype call for one child in SageBio’s rare-disease hackathon. The organisers’ scorer returned a full match.
- Causal-JEPA reproduction2026
An independent reproduction of Causal-JEPA, object-level latent masking. I did not write the paper.
- AIMS: Surgical Phase Detection2025
DINOv2 models for endoscopic surgical workflow at AIMS, later an ISBI 2026 paper.
- MSc Thesis: Surgical Video Prediction2025
A VAE–Transformer for surgical video prediction: +2.36 dB PSNR over the chosen baselines, at 22 FPS.
- Multiview diffusion 3D2024
A public Hugging Face Space for multiview diffusion. The figure is a generated illustration, not a capture of a run.
- Pothole Detection (arXiv)2024
Co-author of a YOLOv7 + ESRGAN preprint: 94.7% precision and 82.6% recall on the PNW table.
- JAX Diffusers Sprint2023
Eighth globally in the 2023 Hugging Face JAX Diffusers sprint, with a ControlNet trained on TPU v4.
- Hopfield Networks & TSP2022
A 2022 sole-author arXiv note on Hopfield networks and simulated annealing for TSP.
- Gemma-Le: VLA Policy—
A compact vision-language-action policy in a LeRobot fork, with checkpoints on the Hub.
- LGM: 3D Model Generation—
A Hugging Face Space that wraps LGM for image-to-3D.
- Deep RL & Hugging Face—
PPO, DQN, Doom, and PyBullet agents from the Hugging Face Deep RL course.