There exists a large number of datasets for organ segmentation, which are partially annotated, and sequentially constructed. A typical dataset is constructed at a certain time by curating medical images and annotating the organs of interest. In other words, new datasets with annotations of new organ categories are built over time. To unleash the potential behind these partially labeled, sequentially-constructed datasets, we propose to learn a multi-organ segmentation model through incremental learning (IL). In each IL stage, we lose access to the previous annotations, whose knowledge is assumingly captured by the current model, and gain the access to a new dataset with annotations of new organ categories, from which we learn to update the organ segmentation model to include the new organs. We give the first attempt to conjecture that the different distribution is the key reason for 'catastrophic forgetting' that commonly exists in IL methods, and verify that IL has the natural adaptability to medical image scenarios. Extensive experiments on five open-sourced datasets are conducted to prove the effectiveness of our method and the conjecture mentioned above.
Feature extraction with convolutional neural networks (CNNs) is a popular method to represent images for machine learning tasks. These representations seek to capture global image content, and ideally should be independent of geometric transformations. We focus on measuring and visualizing the shift invariance of extracted features from popular off-the-shelf CNN models. We present the results of three experiments comparing representations of millions of images with exhaustively shifted objects, examining both local invariance (within a few pixels) and global invariance (across the image frame). We conclude that features extracted from popular networks are not globally invariant, and that biases and artifacts exist within this variance. Additionally, we determine that anti-aliased models significantly improve local invariance but do not impact global invariance. Finally, we provide a code repository for experiment reproduction, as well as a website to interact with our results at https://jakehlee.github.io/visualize-invariance.
Owing to human labor shortages, the automation of labor-intensive manual waste-sorting is needed. The goal of automating the waste-sorting is to replace the human role of robust detection and agile manipulation of the waste items by robots. To achieve this, we propose three methods. First, we propose a combined manipulation method using graspless push-and-drop and pick-and-release manipulation. Second, we propose a robotic system that can automatically collect object images to quickly train a deep neural network model. Third, we propose the method to mitigate the differences in the appearance of target objects from two scenes: one for the dataset collection and the other for waste sorting in a recycling factory. If differences exist, the performance of a trained waste detector could be decreased. We address differences in illumination and background by applying object scaling, histogram matching with histogram equalization, and background synthesis to the source target-object images. Via experiments in an indoor experimental workplace for waste-sorting, we confirmed the proposed methods enable quickly collecting the training image sets for three classes of waste items, i.e., aluminum can, glass bottle, and plastic bottle and detecting them with higher performance than the methods that do not consider the differences. We also confirmed that the proposed method enables the robot quickly manipulate them.
Most of existing image denoising methods assume the corrupted noise to be additive white Gaussian noise (AWGN). However, the realistic noise in real-world noisy images is much more complex than AWGN, and is hard to be modelled by simple analytical distributions. As a result, many state-of-the-art denoising methods in literature become much less effective when applied to real-world noisy images captured by CCD or CMOS cameras. In this paper, we develop a trilateral weighted sparse coding (TWSC) scheme for robust real-world image denoising. Specifically, we introduce three weight matrices into the data and regularisation terms of the sparse coding framework to characterise the statistics of realistic noise and image priors. TWSC can be reformulated as a linear equality-constrained problem and can be solved by the alternating direction method of multipliers. The existence and uniqueness of the solution and convergence of the proposed algorithm are analysed. Extensive experiments demonstrate that the proposed TWSC scheme outperforms state-of-the-art denoising methods on removing realistic noise.
Detecting novel objects from few examples has become an emerging topic in computer vision recently. However, these methods need fully annotated training images to learn new object categories which limits their applicability in real world scenarios such as field robotics. In this work, we propose a probabilistic multiple instance learning approach for few-shot Common Object Localization (COL) and few-shot Weakly Supervised Object Detection (WSOD). In these tasks, only image-level labels, which are much cheaper to acquire, are available. We find that operating on features extracted from the last layer of a pre-trained Faster-RCNN is more effective compared to previous episodic learning based few-shot COL methods. Our model simultaneously learns the distribution of the novel objects and localizes them via expectation-maximization steps. As a probabilistic model, we employ von Mises-Fisher (vMF) distribution which captures the semantic information better than Gaussian distribution when applied to the pre-trained embedding space. When the novel objects are localized, we utilize them to learn a linear appearance model to detect novel classes in new images. Our extensive experiments show that the proposed method, despite being simple, outperforms strong baselines in few-shot COL and WSOD, as well as large-scale WSOD tasks.
The Dice score and Jaccard index are commonly used metrics for the evaluation of segmentation tasks in medical imaging. Convolutional neural networks trained for image segmentation tasks are usually optimized for (weighted) cross-entropy. This introduces an adverse discrepancy between the learning optimization objective (the loss) and the end target metric. Recent works in computer vision have proposed soft surrogates to alleviate this discrepancy and directly optimize the desired metric, either through relaxations (soft-Dice, soft-Jaccard) or submodular optimization (Lov\'asz-softmax). The aim of this study is two-fold. First, we investigate the theoretical differences in a risk minimization framework and question the existence of a weighted cross-entropy loss with weights theoretically optimized to surrogate Dice or Jaccard. Second, we empirically investigate the behavior of the aforementioned loss functions w.r.t. evaluation with Dice score and Jaccard index on five medical segmentation tasks. Through the application of relative approximation bounds, we show that all surrogates are equivalent up to a multiplicative factor, and that no optimal weighting of cross-entropy exists to approximate Dice or Jaccard measures. We validate these findings empirically and show that, while it is important to opt for one of the target metric surrogates rather than a cross-entropy-based loss, the choice of the surrogate does not make a statistical difference on a wide range of medical segmentation tasks.
The compressed sensing (CS) theory has been successfully applied to image compression in the past few years as most image signals are sparse in a certain domain. Several CS reconstruction models have been recently proposed and obtained superior performance. However, there still exist two important challenges within the CS theory. The first one is how to design a sampling mechanism to achieve an optimal sampling efficiency, and the second one is how to perform the reconstruction to get the highest quality to achieve an optimal signal recovery. In this paper, we try to deal with these two problems with a deep network. First of all, we train a sampling matrix via the network training instead of using a traditional manually designed one, which is much appropriate for our deep network based reconstruct process. Then, we propose a deep network to recover the image, which imitates traditional compressed sensing reconstruction processes. Experimental results demonstrate that our deep networks based CS reconstruction method offers a very significant quality improvement compared against state of the art ones.
Lighting estimation from a single image is an essential yet challenging task in computer vision and computer graphics. Existing works estimate lighting by regressing representative illumination parameters or generating illumination maps directly. However, these methods often suffer from poor accuracy and generalization. This paper presents Geometric Mover's Light (GMLight), a lighting estimation framework that employs a regression network and a generative projector for effective illumination estimation. We parameterize illumination scenes in terms of the geometric light distribution, light intensity, ambient term, and auxiliary depth, and estimate them as a pure regression task. Inspired by the earth mover's distance, we design a novel geometric mover's loss to guide the accurate regression of light distribution parameters. With the estimated lighting parameters, the generative projector synthesizes panoramic illumination maps with realistic appearance and frequency. Extensive experiments show that GMLight achieves accurate illumination estimation and superior fidelity in relighting for 3D object insertion.
Image is usually taken for expressing some kinds of emotions or purposes, such as love, celebrating Christmas. There is another better way that combines the image and relevant song to amplify the expression, which has drawn much attention in the social network recently. Hence, the automatic selection of songs should be expected. In this paper, we propose to retrieve semantic relevant songs just by an image query, which is named as the image2song problem. Motivated by the requirements of establishing correlation in semantic/content, we build a semantic-based song retrieval framework, which learns the correlation between image content and lyric words. This model uses a convolutional neural network to generate rich tags from image regions, a recurrent neural network to model lyric, and then establishes correlation via a multi-layer perceptron. To reduce the content gap between image and lyric, we propose to make the lyric modeling focus on the main image content via a tag attention. We collect a dataset from the social-sharing multimodal data to study the proposed problem, which consists of (image, music clip, lyric) triplets. We demonstrate that our proposed model shows noticeable results in the image2song retrieval task and provides suitable songs. Besides, the song2image task is also performed.
State of the art methods in astronomical image reconstruction rely on the resolution of a regularized or constrained optimization problem. Solving this problem can be computationally intensive and usually leads to a quadratic or at least superlinear complexity w.r.t. the number of pixels in the image. We investigate in this work the use of convolutional neural networks for image reconstruction in astronomy. With neural networks, the computationally intensive tasks is the training step, but the prediction step has a fixed complexity per pixel, i.e. a linear complexity. Numerical experiments show that our approach is both computationally efficient and competitive with other state of the art methods in addition to being interpretable.