Grounding Visual Explanations

By Lisa Anne Hendricks, Ronghang Hu, Trevor Darrell, Zeynep Akata. 2018

241 citations recorded by the source corpus.

Existing visual explanation generating agents learn to fluently justify a class prediction. However, they may mention visual attributes which reflect a strong class prior, although the evidence may not actually be in the image. This is particularly concerning as ultimately such agents fail in building trust with human users. To overcome this limitation, we propose a phrase-critic model to refine generated candidate explanations augmented with flipped phrases which we use as negative examples while training. At inference time, our phrase-critic model takes an image and a candidate explanation as input and outputs a score indicating how well the candidate explanation is grounded in the image. Our explainable AI agent is capable of providing counter arguments for an alternative prediction, i.e. counterfactuals, along with explanations that justify the correct classification decisions. Our model improves the textual explanation quality of fine-grained classification decisions on the CUB dataset by mentioning phrases that are grounded in the image. Moreover, on the FOIL tasks, our agent detects when there is a mistake in the sentence, grounds the incorrect phrase and corrects it significantly better than other models.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Multimodal Learning, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Multimodal Learning

In Safety, Robustness & Interpretability

Citations in the corpus

Shaping Visual Representations with Language for Few-Shot Classification (2019, neutral): “…generate natural language explanations of decisions for interpretability for both textual (e.g. natural language inference; [3]) and visual [17, 18] tasks, but here we examine whether this act of predicting language can actually improve downstream task performance; similar ideas have been…”
From Recognition to Cognition: Visual Commonsense Reasoning (2018, background): “This has motivated work in having models provide explanations for their behavior, in the form of a natural language sentence [31, 9, 41] or an attention map [32, 35, 37].”
Generative Counterfactual Introspection for Explainable Deep Learning (2019, methodology): “With the pressing need to obtain causal understanding of model behavior, interpretation approaches [20]–[23] focusing on counterfactual reasoning have been proposed.”
Learning to Explain With Complemental Examples (2018, background): “Few works have treated multimodality for explanation [21, 2], which is visual and linguistic.”
Textual Explanations for Self-Driving Vehicles (2018, methodology): “For an image classification problem, [7,8] used an LSTM [9] caption generation model that generates textual justifications for a CNN model.”
The challenge of crafting intelligible intelligence (2018, background): “ny examples, the explanations sometimes include descriptions of details that aren’t actually present in the image; newer approaches, such as phrase-critic methods, may create even better descriptions [16]. Another approach might determine if there are hidden layers in the learned classifier, that le”
XOC: Explainable Observer-Classifier for Explainable Binary Decisions (2019, background): “, a language model explaining the classification decision of a vision model by mentioning discriminative features of the object [11, 12].”
Counterfactual Visual Explanations (2019, methodology): “(Hendricks et al., 2018) learn a model to generate a counter-factual explanation for why a model predicted a class c instead of class c′, similar to our approach.”
Compressing Visual-linguistic Model via Knowledge Distillation (2021, background): “, image/video captioning [3, 79, 70, 76, 26, 15], VQA [21, 4, 19], textual grounding [56, 16, 24, 82], etc.”
DARPA's Explainable Artificial Intelligence (XAI) Program (2019, methodology): “For data analytics, DEXAI will be demonstrated using visual question answering (VQA) and ﬁ ltering tasks (for example, using large-scale data sets such as VQA-X and ACT-X for VQA tasks and activity recognition tasks, respectively), xView, and Distinct Describable Moments (Hendricks et al. 2018).”
Formalizing Trust in Artificial Intelligence: Prerequisites, Causes and Goals of Human Trust in AI (2020, background): “Transparency High-standard documentation All Technical explainability Factsheets (explainability) Saliency maps [61], self-attention patterns [40], influence functions [38], probing [16] Adaptable user-centered explainability Factsheets (explainability) Counterfactual [21], contrastive [51], natural”
REVERIE: Remote Embodied Visual Referring Expression in Real Indoor Environments (2019, background): “instructions that refer to previously unseen objects in real environments. It represents an extension of existing Referring Expression, Visual Grounding, and Embodied Question Answering datasets (see [6, 8, 9, 13, 29] for example) to the case where it cannot be assumed that the object in question is”
Distilling Model Failures as Directions in Latent Space (2022, unknown): “…et al., 2013; Dabkowski & Gal, 2017; Sundararajan et al., 2017) and perturbation-based (Ribeiro et al., 2016a; Goyal et al., 2019; Fong & Vedaldi, 2017; Dabkowski & Gal, 2017; Zintgraf et al., 2017; Dhurandhar et al., 2018; Chang et al., 2019; Hendricks et al., 2018; Singla et al., 2021) methods.”
Leakage-Adjusted Simulatability: Can Models Generate Non-Trivial Explanations of Their Behavior in Natural Language? (2020, methodology): “, through various feature importance estimates (Hendricks et al., 2018; Ribeiro et al., 2016) or model-generated natural language (NL) (Hendricks et al.”
Explaining NLP Models via Minimal Contrastive Editing (MiCE) (2020, background): “The problem of generating minimal contrastive edits, also called counterfactual explanations (Wachter et al., 2017), 12 has previously been explored for tabular data (Karimi et al., 2020) and images (Hendricks et al., 2018; Goyal et al., 2019; Looveren and Klaise, 2019) but less for language.”
Explainable Artificial Intelligence: a Systematic Review (2020, background): “Relevance and Discriminative Loss [259, 137] generates textual explanations for an image classifier like “The bird in the photo is a White Pelican because.”
Counterfactual VQA: A Cause-Effect Look at Language Bias (2020, background): “Counterfactual thinking and causal inference have inspired several studies in computer vision, including visual explanations [20, 7, 42, 48], scene graph generation [12, 38], image recognition [38], video analysis [16, 26], zero-shot and few-shot learning [50, 49] incremental learning [22], represen”