Human Attention in Visual Question Answering: Do Humans and Deep Networks look at the same regions?
By Abhishek Das, Harsh Agrawal, C. Lawrence Zitnick, Devi Parikh, Dhruv Batra. 2016
500 citations recorded by the source corpus.
We conduct large-scale studies on `human attention' in Visual Question Answering (VQA) to understand where humans choose to look to answer questions about images. We design and test multiple game-inspired novel attention-annotation interfaces that require the subject to sharpen regions of a blurred image to answer a question. Thus, we introduce the VQA-HAT (Human ATtention) dataset. We evaluate attention maps generated by state-of-the-art VQA models against human attention both qualitatively (via visualizations) and quantitatively (via rank-order correlation). Overall, our experiments show that current attention models in VQA do not seem to be looking at the same regions as humans.
Subfields: Computer Vision and Pattern Recognition, Multimodal Learning, Question Answering & Reading Comprehension
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Trevor Darrell
- Xiaogang Wang
- Jitendra Malik
- Cordelia Schmid
- Bernt Schiele
In Multimodal Learning
In Question Answering & Reading Comprehension
- Christopher D. Manning
- Drew A. Hudson
- Mohit Bansal
- Nazneen Fatema Rajani and Raymond J. Mooney
- Christopher Kanan
- Devi Parikh
- Kushal Kafle
- Licheng Yu
Citations in the corpus
- e-SNLI: Natural Language Inference with Natural Language Explanations (2018, neutral): “Similar to the evaluation performed for visual question answering in Das et al. [8], our highlighted words could provide a source of supervision and evaluation for attention models [25, 22] or post-hoc explanation models where the explanation consists of a subset of the input.”
- GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering (2019, background): “VQA models, which become heavily reliant upon such statistical biases and tendencies within the answer distribution to largely circumvent the need for true visual scene understanding [1, 9, 13, 7].”
- Hierarchical Question-Image Co-Attention for Visual Question Answering (2016, methodology): “concurrent work, [5] collected ‘human attention maps’ that are used to evaluate the attention maps generated by attention models for VQA.”
- Embodied Question Answering (2017, background): “– these models often fail to associate entities in text with corresponding image pixels, relying instead on dataset biases to respond seemingly intelligently even when attending to irrelevant regions [2,3]. In EmbodiedQA, we take a goal-driven view of grounding – our agent grounds a visual question ”
- Explaining Explanations: An Overview of Interpretability of Machine Learning (2018, background): “Datasets of human attention have been created [58], [59]; these allow systems to be evaluated according to how closely and their internal attention resembles human attention.”
- Recent advances in convolutional neural networks (2015, methodology): “P ∈R After obtaining the distribution P ( p 1 , ..., p |R j | ) , we can sample from the multi-nomial distribution based on p to pick a location l within the region, then set the pooled activation as y j = a l , where l ∼ P ( p 1 , ..., p |R j | ) .”
- Stacking with Auxiliary Features for Visual Question Answering (2018, methodology): “The rank correlation protocol has been used in the past to compare machinegenerated and human attention-maps as described by Das et al. (2016). We also experimented with using the Earth Mover’s Distance (EMD) in place of the rank-order correlation metric, as discussed in Section 6.”
- Global-and-local attention networks for visual recognition (2018, background): “This can be alleviated in one-player games [26, 34] where a single player may be asked to query a blurred image to answer questions about it by sharpening selected regions.”
- Towards Transparent AI Systems: Interpreting Visual Question Answering Models (2016, methodology): “(Das et al., 2016) recently collected human attention annotations for (question, image) pairs from VQA dataset (Antol et al.”
- Unmasking Clever Hans predictors and assessing what machines really learn (2019, methodology): “Explanation methods were successfully applied to a wide set of complex real-world problems such as the analysis of faces [81, 123, 124], EEG and fMRI data [23, 50], human gait [48], videos [125], speech [126], reinforcement learning [25], biological data [127], text [26,128,129], or comparing human ”
- Neural Baby Talk (2018, background): “They often tend to ‘look’ at different regions than humans would and tend to copy captions from training data [8].”
- TVQA+: Spatio-Temporal Grounding for Video Question Answering (2019, background): “However, sometimes the attention learned by the model itself may not accord with human expectations [22, 5].”