All Open Source Projects

SEM

A post-hoc, zero-shot debiasing framework that operates in a Sparse Autoencoder (SAE) latent space to pinpoint and modulate bias-relevant neurons in Vision-Language Models.

PyTorch Mechanistic Interpretability VLMs Debiasing

[GitHub] [Project Page] [arXiv]

C2B

An unsupervised, task-agnostic framework for discovering biases in pretrained visual classifiers using LLM-generated bias proposals and text-to-image retrieval.

Bias Discovery LLMs VQA Computer Vision

[GitHub] [Video] [arXiv]

DVMS

A PyTorch-based Deep Variational Learning framework for multiple trajectory prediction of 360° head movements.

PyTorch Trajectory Prediction VR/360° Deep Learning

[GitLab] [Paper]

SMART360

A 360° streaming simulation environment and preprocessing pipeline to compare motion prediction and adaptive bitrate (ABR) strategies in virtual reality.

Simulation Adaptive Streaming VR/360° Docker

[Simulator] [Preprocessing] [Paper]

PEM360

A processing suite and dataset of 360° videos with continuous physiological measurements, subjective emotional ratings, and motion traces. Includes tools for computing saliency and emotional maps.

Dataset Physiological Data Saliency Maps Python

[GitLab] [Paper]