Simple Baseline for Visual Question Answering

About

We describe a very simple bag-of-words baseline for visual question answering. This baseline concatenates the word features from the question and CNN features from the image to predict the answer. When evaluated on the challenging VQA dataset [2], it shows comparable performance to many recent approaches using recurrent neural networks. To explore the strength and weakness of the trained model, we also provide an interactive web demo and open-source code. .

Bolei Zhou, Yuandong Tian, Sainbayar Sukhbaatar, Arthur Szlam, Rob Fergus · 2015

Related benchmarks

Task	Dataset	Result
Visual Question Answering	VQA (test-dev)	Acc (All)55.72	147
Visual Question Answering	VQA (test-std)	Accuracy—	120
Open-Ended Visual Question Answering	VQA 1.0 (test-dev)	Overall Accuracy55.72	100
Visual Question Answering (Multiple-choice)	VQA 1.0 (test-dev)	Accuracy (All)61.7	66
Visual Question Answering	CLEVR (test)	Overall Accuracy48.4	61
Open-Ended Visual Question Answering	VQA 1.0 (test-standard)	Overall Accuracy55.9	50
Open-Ended Visual Question Answering	VQA (test-standard)	Accuracy (Overall)55.9	32
Visual Question Answering	VQA 1 (test-standard)	VQA Open-Ended Accuracy (All)55.89	28
Visual Question Answering (Multiple-choice)	VQA 1.0 (test-standard)	Accuracy (All)62	27
Visual Question Answering (Multiple-choice)	VQA (test-standard)	Accuracy (All)62	17