An alternating RL/SL training scheme for GuessWhich, where the RL policy acts on image guesses rather than words, achieves higher guessing accuracy and lower perplexity than word-level RL baselines.
GuessWhat?! Visual object discovery through multi-modal dialogue
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We introduce GuessWhat?!, a two-player guessing game as a testbed for research on the interplay of computer vision and dialogue systems. The goal of the game is to locate an unknown object in a rich image scene by asking a sequence of questions. Higher-level image understanding, like spatial reasoning and language grounding, is required to solve the proposed task. Our key contribution is the collection of a large-scale dataset consisting of 150K human-played games with a total of 800K visual question-answer pairs on 66K images. We explain our design decisions in collecting the dataset and introduce the oracle and questioner tasks that are associated with the two players of the game. We prototyped deep learning models to establish initial baselines of the introduced tasks.
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Building Task-Oriented Visual Dialog Systems Through Alternative Optimization Between Dialog Policy and Language Generation
An alternating RL/SL training scheme for GuessWhich, where the RL policy acts on image guesses rather than words, achieves higher guessing accuracy and lower perplexity than word-level RL baselines.