A visual Turing test for computer vision systems

By Donald Geman, Stuart Geman, Neil Hallonquist, Laurent Younes. 2014

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes (2020, background): “Many practitioners believe that multimodality holds the key to problems as varied as natural language understanding [5, 41], computer vision evaluation [23], and embodied AI [64].”
Learning Visual N-Grams from Web Data (2016, methodology): “Finally, we aim to use our models in tasks that require recognition of a large variety of visual concepts and relations between them, such as visual question answering [2, 67], visual Turing tests [21], and scene graph prediction [36].”
Visual Dialog (2016, methodology): “[13] is a statistical templated-question generator and not an actual visual dialog system.”
VQA: Visual Question Answering (2015, background): “Critical to achieving success on this more difficult and unconstrained task, our VQA dataset is two orders of magnitude larger than [16, 30] (>250,000 vs.”
FiLM: Visual Reasoning with a General Conditioning Layer (2017, background): “Visual question answering itself has its own line of datasets (Malinowski and Fritz 2014; Geman et al. 2015; Antol et al. 2015) which focus on asking a diverse set of simpler questions on images, often answerable in a single glance.”
Neural Module Networks (2015, background): “Visual Question Answering Answering questions about images is sometimes referred to as a “Visual Turing Test” [27, 11].”
Visual7W: Grounded Question Answering in Images (2015, methodology): “’s proposal [8] and VQA’s approach [1].”
Are You Talking to a Machine? Dataset and Methods for Multilingual Image Question (2015, background): “, there are only 2591 and 1449 images for [9] and [22] respectively).”
Ask Your Neurons: A Neural-Based Approach to Answering Questions about Images (2015, background): “[4] proposes such an approach where the task is limited to yes/no answers, and [24] proposes a set of different question answering benchmarks us-ing synthetic data.”
Dynamic Memory Networks for Visual and Textual Question Answering (2016, background): “While VQA datasets existed before they did not include open-ended, free-form questions about general images (Geman et al., 2014).”
Yin and Yang: Balancing and Answering Binary Visual Questions (2015, background): “For example, [19] proposes a system to generate binary questions from templates using a fixed vocabulary of objects, attributes, and relationships between objects.”
Visual Madlibs: Fill in the blank Image Generation and Question Answering (2015, methodology): “Geman et al. [12] design a visual Turing test to test image understanding using a series of binary questions about image content.”
Generation and Comprehension of Unambiguous Object Descriptions (2015, methodology): “Referring expressions is re-lated to the task of VQA (see e.g., [2, 38, 39, 16, 15]).”
Ask Me Anything: Free-Form Visual Question Answering Based on Knowledge from External Sources (2015, background): “[12] proposed an automatic ‘query generator’ that was trained on annotated images and produced a sequence of binary questions from any given test image.”
GuessWhat?! Visual Object Discovery through Multi-modal Dialogue (2016, background): “Recently, the VQA challenge [6] has provided a new dataset far bigger than previous attempts [13, 27] where, much like in GuessWhat?!, questions are free-form.”