Learning to Refine Object Segments

Object segmentation requires both object-level information and low-level pixel data. This presents a challenge for feedforward networks: lower layers in convolutional nets capture rich spatial information, while upper layers encode object-level knowledge but are invariant to factors such as pose and appearance. In this work we propose to augment feedforward nets for object segmentation with a novel top-down refinement approach. The resulting bottom-up/top-down architecture is capable of efficiently generating high-fidelity object masks. Similarly to skip connections, our approach leverages features at all layers of the net. Unlike skip connections, our approach does not attempt to output independent predictions at each layer. Instead, we first output a coarse ‘mask encoding’ in a feedforward pass, then refine this mask encoding in a top-down pass utilizing features at successively lower layers. The approach is simple, fast, and effective. Building on the recent DeepMask network for generating object proposals, we show accuracy improvements of 10–20% in average recall for various setups. Additionally, by optimizing the overall network architecture, our approach, which we call SharpMask, is 50 % faster than the original DeepMask network (under .8 s per image).

Weakly Supervised Object Segmentation with 004 dwaeConvolutional Neural Networks

Ronan Collobert, Pedro Henrique Oliveira Pinheiro

Can a machine learn how to segment different objects in real world images without having any prior knowledge about the delineation of the classes? In this paper, we demonstrate that this task is indeed possible. We address the problem by training a Convolutional Neural Networks (CNN) model with weakly labeled images, \emph{i.e.}, images in which the only knowledge assumed on each sample is the presence or not of an object. The model, trained in an one--vs-all scheme, learns representations that distinguish image patches that belong to the class of interest from those that belong to the background. The per-pixel segmentation is obtained by applying the model to the patch surrounding the pixel and assigning the inferred class to that pixel. Our system is trained using a subset of the Imagenet dataset. The experiments are validated on two challenging classes for segmentation: cats and dogs. We show both quantitatively and qualitatively that the model achieves good accuracy results for these classes on the Pascal VOC 2012 competition, without using any prior segmentation knowledge. This model is powerful in the sense that it learns how to segment objects without the use of costly fully-labeled segmentation datasets.

Idiap2014

DermoNet: Densely Linked Convolutional Neural Network for Efficient Skin Lesion Segmentation

Seyedbehzad Bozorgtabar, Jean-Philippe Thiran

Recent state-of-the-art methods for skin lesion segmentation are based on Convolutional Neural Networks (CNNs). Even though these CNN based segmentation approaches are accurate, they are computationally expensive. In this paper, we address this problem and propose an efficient fully Convolutional Neural Network, named DermoNet. In DermoNet, due to our densely connected convolutional blocks and skip connections, network layers can reuse information from their preceding layers and ensure high accuracy in later network layers. By doing so, we take advantage of the capability of high-level feature representations learned at intermediate layers with varying scales and resolutions for lesion segmentation. Quantitative evaluation is conducted on three well-established public benchmark datasets: the ISBI 2016, ISBI 2017, and the PH2 datasets. The experimental results show that our proposed approach outperforms state-of-the-art algorithms on these three datasets. We also compared the runtime performance of DermoNet with two other related architectures, which are FCN and U-Net. The proposed approach is found to be faster and suitable for practical applications.

2018

DermoNet: densely linked convolutional neural network for efficient skin lesion segmentation

Saleh Baghersalimi, Hazim Kemal Ekenel, Jean-Philippe Thiran

Recent state-of-the-art methods for skin lesion segmentation are based on convolutional neural networks (CNNs). Even though these CNN-based segmentation approaches are accurate, they are computationally expensive. In this paper, we address this problem and propose an efficient fully convolutional neural network, named DermoNet. In DermoNet, due to our densely connected convolutional blocks and skip connections, network layers can reuse information from their preceding layers and ensure high accuracy in later network layers. By doing so, we take advantage of the capability of high-level feature representations learned at intermediate layers with varying scales and resolutions for lesion segmentation. Quantitative evaluation is conducted on three well-established public benchmark datasets: the ISBI 2016, ISBI 2017, and the PH2 datasets. The experimental results show that our proposed approach outperforms the state-of-the-art algorithms on these three datasets. We also compared the runtime performance of DermoNet with two other related architectures, which are fully convolutional networks and U-Net. The proposed approach is found to be faster and suitable for practical applications.

2019