CuriosAI Submission to the EgoExo4D Proficiency Estimation Challenge 2025

Hayato Tanoue Hiroki Nishihara Yuma Suzuki Takayuki Hori Hiroki Takushima Aiswariya Manoj Kumar Yuki Shibata Mitsuru Takeda Fumika Beppu Zhao Hengwei Yuto Kanda Daichi Yamaga SoftBank Corp. AI & Data Technology Planning Division {hayato.tanoue, hiroki.nishihara, yuma.suzuki, takayuki.hori, hiroki.ta · 2025

Read CuriosAI Submission to the EgoExo4D Proficiency Estimation Challenge 2025 by Hayato Tanoue Hiroki Nishihara Yuma Suzuki Takayuki Hori Hiroki Takushima Aiswariya Manoj Kumar Yuki Shibata Mitsuru Takeda Fumika Beppu Zhao Hengwei Yuto Kanda Daichi Yamaga SoftBank Corp. AI & Data Technology Planning Division {hayato.tanoue, hiroki.nishihara, yuma.suzuki, takayuki.hori, hiroki.ta (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-08022.txt.

Description

Hayato Tanoue and co-authors report their entry to a challenge on judging how skilled a person is from video. The EgoExo4D challenge supplies recordings from body-worn and external cameras, and systems must estimate the performer's proficiency. The team's report compares two designs. One trains a large vision model, Sapiens-2B, to predict skill level and scenario together in a single multi-task setup.

The second design works in two stages: first recognise the scenario with a zero-shot method, then apply skill classifiers tuned to each camera view. That pipeline scored 47.8 percent accuracy, beating the joint model's 43.6 percent. The authors take the result as evidence that conditioning skill judgements on the scenario, knowing what activity is being performed before grading it, is the stronger recipe for this task.

Catalog record

CREATORHayato Tanoue Hiroki Nishihara Yuma Suzuki Takayuki Hori Hiroki Takushima Aiswariya Manoj Kumar Yuki Shibata Mitsuru Takeda Fumika Beppu Zhao Hengwei Yuto Kanda Daichi Yamaga SoftBank Corp. AI & Data Technology Planning Division {hayato.tanoue, hiroki.nishihara, yuma.suzuki, takayuki.hori, hiroki.ta
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE9 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-08022

Browse the full library · About the Tower of Babel