To tackle increasingly complex tasks, it has become an essential ability of neural networks to learn abstract representations. These task-specific representations and, particularly, the invariances they capture turn neural networks into black box models that lack interpretability. To open such a black box, it is, therefore, crucial to uncover the different semantic concepts a model has learned as well as those that it has learned to be invariant to. We present an approach based on INNs that (i) recovers the task-specific, learned invariances by disentangling the remaining factor of variation in the data and that (ii) invertibly transforms these recovered invariances combined with the model representation into an equally expressive one with accessible semantic concepts. As a consequence, neural network representations become understandable by providing the means to (i) expose their semantic meaning, (ii) semantically modify a representation, and (iii) visualize individual learned semantic concepts and invariances. Our invertible approach significantly extends the abilities to understand black box models by enabling post-hoc interpretations of state-of-the-art networks without compromising their performance.
论文检索
输入标题、作者或关键词，从 1,526 篇学术成果中精准定位
We consider universal adversarial patches for faces --- small visual elements whose addition to a face image reliably destroys the performance of face detectors. Unlike previous work that mostly focused on the algorithmic design of adversarial examples in terms of improving the success rate as an attacker, in this work we show an interpretation of such patches that can prevent the state-of-the-art face detectors from detecting the real faces. We investigate a phenomenon: patches designed to suppress real face detection appear face-like. This phenomenon holds generally across different initialization, locations, scales of patches, backbones, and state-of-the-art face detection frameworks. We propose new optimization-based approaches to automatic design of universal adversarial patches for varying goals of the attack, including scenarios in which true positives are suppressed without introducing false positives. Our proposed algorithms perform well on real-world datasets, deceiving state-of-the-art face detectors in terms of multiple precision/recall metrics and transferability.
Visual reasoning is crucial for visual question answering (VQA). However, without labelled programs, implicit reasoning under natural supervision is still quite challenging and previous models are hard to interpret. In this paper, we rethink implicit reasoning process in VQA, and propose a new formulation which maximizes the log-likelihood of joint distribution for the observed question and predicted answer. Accordingly, we derive a Temporal Reasoning Network (TRN) framework which models the implicit reasoning process as sequential planning in latent space. Our model is interpretable on both model design in probabilist and reasoning process via visualization. We experimentally demonstrate that TRN can support implicit reasoning across various datasets. The experiment results of our model are competitive to existing implicit reasoning models and surpass baseline by large margin on complicated reasoning tasks without extra computation cost in forward stage.
Recent Transformer-based large-scale pre-trained models have revolutionized vision-and-language (V+L) research. Models such as ViLBERT, LXMERT and UNITER have significantly lifted state of the art across a wide range of V+L benchmarks. However, little is known about the inner mechanisms that destine their impressive success. To reveal the secrets behind the scene, we present VALUE (Vision-And-Language Understanding Evaluation), a set of meticulously designed probing tasks (e.g., Visual Coreference Resolution, Visual Relation Detection) generalizable to standard pre-trained V+L models, to decipher the inner workings of multimodal pre-training (e.g., implicit knowledge garnered in individual attention heads, inherent cross-modal alignment learned through contextualized multimodal embeddings). Through extensive analysis of each archetypal model architecture via these probing tasks, our key observations are: (i) Pre-trained models exhibit a propensity for attending over text rather than images during inference. (ii) There exists a subset of attention heads that are tailored for capturing cross-modal interactions. (iii) Learned attention matrix in pre-trained models demonstrates patterns coherent with the latent alignment between image regions and textual words. (iv) Plotted attention patterns reveal visually-interpretable relations among image regions. (v) Pure linguistic knowledge is also effectively encoded in the attention heads. These are valuable insights serving to guide future work towards designing better model architecture and objectives for multimodal pre-training.
To what extent are two images picturing the same 3D surfaces? Even when this is a known scene, the answer typically requires an expensive search across scale space, with matching and geometric verification of large sets of local features. This expense is further multiplied when a query image is evaluated against a gallery, e.g. in visual relocalization. While we don't obviate the need for geometric verification, we propose an interpretable image-embedding that cuts the search in scale space to essentially a lookup.Our approach measures the asymmetric distance between two images. The model then learns a scene-specific measure of similarity, from training examples with known 3D visible-surface overlaps. The result is that we can quickly identify, for example, which test image is a close-up version of another, and by what scale factor. Subsequently, local features need only be detected at that scale. We validate our scene-specific model by showing how this embedding yields competitive image-matching results, while being simpler, faster, and also interpretable by humans.
Textures in natural images can be characterized by color, shape, periodicity of elements within them, and other attributes that can be described using natural language. In this paper, we study the problem of describing visual attributes of texture on a novel dataset containing rich descriptions of textures, and conduct a systematic study of current generative and discriminative models for grounding language to images on this dataset. We find that while these models capture some properties of texture, they fail to capture several compositional properties, such as the colors of dots. We provide critical analysis of existing models by generating synthetic but realistic textures with different descriptions. Our dataset also allows us to train interpretable models and generate language-based explanations of what discriminative features are learned by deep networks for fine-grained categorization where texture plays a key role. We present visualizations of several fine-grained domains and show that texture attributes learned on our dataset offer improvements over expert-designed attributes on the Caltech-UCSD Birds dataset.