Deep convolutional filter banks for texture recognition and segmentation

By Mircea Cimpoi, Subhransu Maji, Andrea Vedaldi. 2014

55 citations recorded by the source corpus.

Research in texture recognition often concentrates on the problem of material recognition in uncluttered conditions, an assumption rarely met by applications. In this work we conduct a first study of material and describable texture at- tributes recognition in clutter, using a new dataset derived from the OpenSurface texture repository. Motivated by the challenge posed by this problem, we propose a new texture descriptor, D-CNN, obtained by Fisher Vector pooling of a Convolutional Neural Network (CNN) filter bank. D-CNN substantially improves the state-of-the-art in texture, mate- rial and scene recognition. Our approach achieves 82.3% accuracy on Flickr material dataset and 81.1% accuracy on MIT indoor scenes, providing absolute gains of more than 10% over existing approaches. D-CNN easily trans- fers across domains without requiring feature adaptation as for methods that build on the fully-connected layers of CNNs. Furthermore, D-CNN can seamlessly incorporate multi-scale information and describe regions of arbitrary shapes and sizes. Our approach is particularly suited at lo- calizing stuff categories and obtains state-of-the-art re- sults on MSRC segmentation dataset, as well as promising results on recognizing materials and surface attributes in clutter on the OpenSurfaces dataset.

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

Accelerating Very Deep Convolutional Networks for Classification and Detection (2015, background): “The very deep VGG models [1] have substantially 679 improved a wide range of visual recognition tasks, 680 including object detection [9], [2], [10], [11], semantic 681 segmentation [12], [13], [14], [40], [41], image cap- 682 tioning [42], [43], [44], video/action recognition [45], 683 image questi”
Deep Image: Scaling up Image Recognition (2015, methodology): “h as different types of birds and flowers, which is different from generic classification. Much work has been devoted on the learning of object parts that benefits from the alignment and segmentation [1, 8, 6, 18]. Current state-of-the-art results for bird species recognition have been achieved in [”
A Wholistic View of Continual Learning with Deep Neural Networks: Forgotten Lessons and the Bridge to Active and Open World Learning (2020, background): “Because there is an abundance of material types, it has become the de-facto standard to collapse any available imagery that is connected to less important materials or where meager amounts of data are available into a single “other” material (Cimpoi et al., 2015; Bell et al., 2015).”
Texture synthesis and the controlled generation of natural stimuli using convolutional neural networks (2015, methodology): “In particular Cimpoi et al. report impressive performance in material recognition and scene segmentation by using a stationary Fisher-Vector representation built on the highest convolutional layer of readily trained neural networks [2].”
Texture Synthesis Using Convolutional Neural Networks (2015, methodology): “This general strategy has recently been employed to improve performance in object recognition and detection [9] or texture recognition and segmentation [3].”
Convolutional Features for Correlation Filter Based Visual Tracking (2015, background): “A multi-scale convolutional feature based approach is proposed by [6] for texture classification and object recognition.”
Learning the Structure of Deep Architectures Using L1 Regularization (2015, methodology): “DNN methods have indeed resulted in very large increases in performance on a wide range of image classification datasets, including large scale datasets such as ImageNet [10], and, by means of transfer learning, smaller datasets such as PascalVOC 2007/2012 [15], SUN397 [7] and MIT Indoor datasets [3”
Handcrafted vs. non-handcrafted features for computer vision classification (2017, methodology): “Several CNNs were used for extracting a set of feature vectors from an image, which were then used to feed an ensemble of SVMs that were finally combined by sum rule.”
Universal representations: The missing link between faces, text, planktons, and cat breeds (2017, background): “For example, following the common intuition that early layers of a neural networks have are less specialized and hence less domain-speciﬁc [7], φ 0 may contain all the early layers up to some depth, after which the different networks branch off.”
Deep Neural Networks as a Computational Model for Human Shape Sensitivity (2016, methodology): “Small modifications to this architecture, such as adding feature map correlations [52,53] or performing VLAD [54] or Fisher Vector [55] pooling already provides convnets with the ability to segment input images and represent textures and artistic style, all of which might be the part of feedforward ”