Neural Module Networks

CVPR 2016  ·  Jacob Andreas, Marcus Rohrbach, Trevor Darrell, Dan Klein ·

Visual question answering is fundamentally compositional in nature---a question like "where is the dog?" shares substructure with questions like "what color is the dog?".. and "where is the cat?" This paper seeks to simultaneously exploit the representational capacity of deep networks and the compositional linguistic structure of questions. We describe a procedure for constructing and learning *neural module networks*, which compose collections of jointly-trained neural "modules" into deep networks for question answering. Our approach decomposes questions into their linguistic substructures, and uses these structures to dynamically instantiate modular networks (with reusable components for recognizing dogs, classifying colors, etc.). The resulting compound networks are jointly trained. We evaluate our approach on two challenging datasets for visual question answering, achieving state-of-the-art results on both the VQA natural image dataset and a new dataset of complex questions about abstract shapes. read more

PDF Abstract CVPR 2016 PDF CVPR 2016 Abstract

Datasets


Introduced in the Paper:

SHAPES

Used in the Paper:

COCO Visual Question Answering
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Visual Question Answering VQA v1 test-dev NMN+LSTM+FT Accuracy 58.6 # 7
Visual Question Answering VQA v1 test-std NMN+LSTM+FT Accuracy 58.7 # 6

Methods


No methods listed for this paper. Add relevant methods here