Farros Alferro smiling at an event in Japan

Farros Alferro

Computer Vision Researcher & ML Engineer

I build and evaluate visual systems for autonomous driving and open-vocabulary search, focusing on reliability beyond familiar data.

EmailGitHubLinkedInRésumé

Research

Annotated high-resolution cafe scene with small-object boxes alongside native-resolution object crops.

Open-vocabulary visual search

Exploring native-resolution strategies for finding small, open-vocabulary targets in high-resolution images.

  • Open-vocabulary detection
  • MLLMs
  • Benchmarking
CARLA sequence comparing the same driving scenario with a reference sedan, an appearance-shifted SUV, and a geometry-shifted dump truck.

OOD-PerceptionBench

A closed-loop benchmark for how end-to-end driving systems respond to appearance and geometry shifts.

  • Autonomous driving
  • CARLA
  • OOD robustness
Free360 pipeline for scene-graph generation and scene-graph-based question answering over a panoramic image.

360Bench & Free360

A panoramic VQA benchmark and training-free framework for structured reasoning over 360° images.

  • 360° vision
  • VQA
  • Scene graphs
Gecko model diagram showing keyword extraction, token alignment, patch selection, and input to a large language model.

Gecko

A training-free, plug-and-play method that helps vision-language models locate small details in high-resolution images.

  • High-resolution vision
  • Vision-language models
  • Training-free

Work & Education

Experience

  1. Machine Learning Engineer Intern

    Mercari, Inc.

    Apr.–Jun. 2026

  2. Research Assistant

    Okatani Computer Vision Lab, Tohoku University

    Oct. 2024–Mar. 2026

  3. Machine Learning Engineer Intern

    Kisuitech

    Jul.–Aug. 2024

Education

  1. M.Sc. in Information Sciences

    Tohoku University

    Oct. 2024–Sep. 2026

  2. B.Eng. in Mechanical & Aerospace Engineering

    Tohoku University

    Oct. 2020–Sep. 2024

Engineering & Open Source

Fashion Recommender Agent

Built a conversational wardrobe assistant that combines retrieval, recommendations, virtual try-on, and web search behind a tool-using agent.

Python · PyTorch · Hugging Face Transformers · OpenCV · CARLA · FastAPI · Docker