In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network to have remarkable localization ability despite being trained on image-level labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that can be applied to a variety of tasks. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1% top-5 error for object localization on ILSVRC 2014, which is remarkably close to the 34.2% top-5 error achieved by a fully supervised CNN approach. We demonstrate that our network is able to localize the discriminative image regions on a variety of tasks despite not being trained for them.

The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification performance at tasks such as visual object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide scene classification CNNs (Places-CNNs) as baselines, that significantly outperform the previous approaches. Visualization of the CNNs trained on Places shows that object detectors emerge as an intermediate representation of scene classification. With its high-coverage and high-diversity of exemplars, the Places Database along with the Places-CNNs offer a novel resource to guide future progress on scene recognition problems.

Scene categorization is a fundamental problem in computer vision. However, scene understanding research has been constrained by the limited scope of currently-used databases which do not capture the full variety of scene categories. Whereas standard databases for object categorization contain hundreds of different classes of objects, the largest available dataset of scene categories contains only 15 classes. In this paper we propose the extensive Scene UNderstanding (SUN) database that contains 899 categories and 130,519 images. We use 397 well-sampled categories to evaluate numerous state-of-the-art algorithms for scene recognition and establish new bounds of performance. We measure human scene classification performance on the SUN database and compare this with computational methods. Additionally, we study a finer-grained scene representation to detect scenes embedded inside of larger scenes.

Many experiments have shown that the human visual system makes extensive use of contextual information for facilitating object search in natural scenes. However, the question of how to formally model contextual influences is still open. On the basis of a Bayesian framework, the authors present an original approach of attentional guidance by global scene context. The model comprises 2 parallel pathways; one pathway computes local features (saliency) and the other computes global (scenecentered) features. The contextual guidance model of attention combines bottom-up saliency, scene context, and top-down mechanisms at an early stage of visual processing and predicts the image regions likely to be fixated by human observers performing natural search tasks in real-world scenes.Keywords: eye movements, visual search, context, global feature, Bayesian model According to feature-integration theory (Treisman & Gelade, 1980), the search for objects requires slow serial scanning because attention is necessary to integrate low-level features into single objects. Current computational models of visual attention based on saliency maps have been inspired by this approach, as it allows a simple and direct implementation of bottom-up attentional mechanisms that are not task specific. Computational models of image saliency (Itti, Koch, & Niebur, 1998;Koch & Ullman, 1985;Parkhurst, Law, & Niebur, 2002;Rosenholtz, 1999) provide some predictions about which regions are likely to attract observers' attention. These models work best in situations in which the image itself provides little semantic information and in which no specific task is driving the observer's exploration. In real-world images, the semantic content of the scene, the co-occurrence of objects, and task constraints have been shown to play a key role in modulating where attention and eye movements go

We propose a general framework called Network Dissection for quantifying the interpretability of latent representations of CNNs by evaluating the alignment between individual hidden units and a set of semantic concepts. Given any CNN model, the proposed method draws on a broad data set of visual concepts to score the semantics of hidden units at each intermediate convolutional layer. The units with semantics are given labels across a range of objects, parts, scenes, textures, materials, and colors. We use the proposed method to test the hypothesis that interpretability of units is equivalent to random linear combinations of units, then we apply our method to compare the latent representations of various networks when trained to solve different supervised and self-supervised training tasks. We further analyze the effect of training iterations, compare networks trained with different initializations, examine the impact of network depth and width, and measure the effect of dropout and batch normalization on the interpretability of deep visual representations. We demonstrate that the proposed method can shed light on characteristics of CNN models and training methods that go beyond measurements of their discriminative power.