BestHub
Sign in
Meituan Technology Team
Meituan Technology Team
Oct 15, 2020 · Artificial Intelligence

Answer-Driven Visual State Estimator for Goal-Oriented Visual Dialogue

The paper introduces the Answer‑Driven Visual State Estimator (ADVSE), which uses answer‑driven focusing attention and conditional visual information fusion to dynamically incorporate answers into visual dialogue, overcoming static encoding limitations and achieving state‑of‑the‑art performance on the GuessWhat?! question‑generation and guessing tasks.

Attention Mechanismgoal-orientedmultimodal AI
0 likes · 10 min read
Answer-Driven Visual State Estimator for Goal-Oriented Visual Dialogue
Alibaba Cloud Developer
Alibaba Cloud Developer
Jun 28, 2019 · Artificial Intelligence

Alibaba AI Wins Visual Dialogue Challenge with New Recursive Model

In the second Visual Dialogue Challenge, Alibaba’s AI outperformed ten teams—including Microsoft and Seoul University—achieving a 74.57% accuracy, surpassing the previous record by 16.82% and exceeding human performance, thanks to its novel recursive exploration dialogue model that integrates image recognition, relational reasoning, and natural language understanding.

AIcomputer visionnatural language processing
0 likes · 4 min read
Alibaba AI Wins Visual Dialogue Challenge with New Recursive Model