We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

计量经济分析 = Econometric analysis

Both convolutional and recurrent operations are building blocks that process one local neighborhood at a time. In this paper, we present non-local operations as a generic family of building blocks for capturing long-range dependencies. Inspired by the classical non-local means method [4] in computer vision, our non-local operation computes the response at a position as a weighted sum of the features at all positions. This building block can be plugged into many computer vision architectures. On the task of video classification, even without any bells and whistles, our nonlocal models can compete or outperform current competition winners on both Kinetics and Charades datasets. In static image recognition, our non-local models improve object detection/segmentation and pose estimation on the COCO suite of tasks. Code will be made available.

/pdf/non-local-neural-networks-3293fthesb.pdf

Non-local Neural Networks

We propose a novel image denoising strategy based on an enhanced sparse representation in transform domain. The enhancement of the sparsity is achieved by grouping similar 2D image fragments (e.g., blocks) into 3D data arrays which we call "groups." Collaborative Altering is a special procedure developed to deal with these 3D groups. We realize it using the three successive steps: 3D transformation of a group, shrinkage of the transform spectrum, and inverse 3D transformation. The result is a 3D estimate that consists of the jointly filtered grouped image blocks. By attenuating the noise, the collaborative filtering reveals even the finest details shared by grouped blocks and, at the same time, it preserves the essential unique features of each individual block. The filtered blocks are then returned to their original positions. Because these blocks are overlapping, for each pixel, we obtain many different estimates which need to be combined. Aggregation is a particular averaging procedure which is exploited to take advantage of this redundancy. A significant improvement is obtained by a specially developed collaborative Wiener filtering. An algorithm based on this novel denoising strategy and its efficient implementation are presented in full detail; an extension to color-image denoising is also developed. The experimental results demonstrate that this computationally scalable algorithm achieves state-of-the-art denoising performance in terms of both peak signal-to-noise ratio and subjective visual quality.

Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering

The discriminative model learning for image denoising has been recently attracting considerable attentions due to its favorable denoising performance. In this paper, we take one step forward by investigating the construction of feed-forward denoising convolutional neural networks (DnCNNs) to embrace the progress in very deep architecture, learning algorithm, and regularization method into image denoising. Specifically, residual learning and batch normalization are utilized to speed up the training process as well as boost the denoising performance. Different from the existing discriminative denoising models which usually train a specific model for additive white Gaussian noise at a certain noise level, our DnCNN model is able to handle Gaussian denoising with unknown noise level (i.e., blind Gaussian denoising). With the residual learning strategy, DnCNN implicitly removes the latent clean image in the hidden layers. This property motivates us to train a single DnCNN model to tackle with several general image denoising tasks, such as Gaussian denoising, single image super-resolution, and JPEG image deblocking. Our extensive experiments demonstrate that our DnCNN model can not only exhibit high effectiveness in several general image denoising tasks, but also be efficiently implemented by benefiting from GPU computing.

Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising

In this paper we review and discuss some of the theoretical and practical aspects, the problems, and the considerations that pushed our research from the one-dimensional LPA-ICI (local polynomial approximation - interesection of conÞdence intervals) algorithm [27] to the development of powerful transform-based methods for anisotropic image processing [18, 15, 19, 20, 8, 21]. In this paper we do not present a new algorithm. Instead, we propose a different and more general interpretation of our recently-developed image denoising algorithms. In particular, we show how they can be treated as nonparametric estimators based on aggregation of a family of local estimates which are pointwise-adaptive in terms of both shape and order.

From local polynomial approximation to pointwise shape-adaptive transforms: an evolutionary nonparametric regression perspective

This article presents a novel and efficient approach to color in high dynamic range (HDR) imaging. In contrast to state-of-the-art methods, we propose to move the complete HDR imaging process from RGB to a luminance–chrominance color space. Our aim is to get a computationally efficient technique and to avoid any possible color distortions originating from the three RGB color channels processed separately. To achieve this, we build a camera response function for the luminance channel only and weight and compose the HDR luminance accordingly, while for the chrominance channels we apply weighting in relation with the saturation level. We demonstrate that our technique yields natural and pleasant to perceive tone-mapped images and is also more robust to noise. © 2007 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 17, 152–162, 2007

Color high dynamic range imaging: The luminance–chrominance approach

BM3D-HVS: Content-adaptive denoising for improved visual quality.

Benchmarking 6D object pose estimation for robotics is not straightforward as sufficient accuracy depends on many factors, e.g., the selected gripper, dimensions, weight and material of an object, grasping point, and the robot task itself. We formulate the problem as a “successful grasp”, i.e. for a fixed set of factors affecting the task, will the given pose estimate provide sufficiently good grasp to complete the task. The successful grasp is modelled in a probabilistic framework by sampling in the pose error space and executing the task and automatically detecting success or failure. Hours of sampling and thousands of samples are used to construct a non-parametric probability of a successful grasp given the pose residual. The framework is experimentally validated with real objects and assembly tasks and comparison of several state-ofthe-art point cloud based 3D pose estimation methods.

Benchmarking 6D Object Pose Estimation for Robotics.

A novel deconvolution technique for blurred observations corrupted by signal-dependent noise is presented. Deblurring is performed with a transform-domain inverse-Þltering applied locally, on a sliding block of adaptively-selected pointwise varying size. Simulation results demonstrate a good quality of the proposed method, which is versatile and which can be easily combined with other transform-domain processing. 1. INTRODUCTION AND MOTIVATION

Alessandro Foi

Papers

From local polynomial approximation to pointwise shape-adaptive transforms: an evolutionary nonparametric regression perspective

Color high dynamic range imaging: The luminance–chrominance approach

BM3D-HVS: Content-adaptive denoising for improved visual quality.

Benchmarking 6D Object Pose Estimation for Robotics.

Adaptive-size block transforms for poissonian image deblurring