Target Objects Of Interest Research Articles

Deep interactive image segmentation based on region and Boundary-click guidance

In interactive image segmentation, the target object of interest can be extracted based on the guidance of user interactions. One of the main goals in this task is to reduce the user interaction burden and ensure satisfactory segmentation with as few interactions as possible. Thanks to the development of deep learning technology, neural network-based interactive approaches have significantly improved the segmentation performance through powerful feature representation. Only limited point (click) interaction is required for the user to complete the segmentation. This paper mainly follows the deep learning-based interactive segmentation methods and explores more efficient interaction strategies and effective segmentation models. We further simplify user interaction to two clicks, where the first click is utilized to select the target region and the other aims to determine the target boundary. Based on the region and boundary clicks, an interactive two-stream network structure is naturally derived to learn the region and boundary features of interest. Furthermore, we also construct an information diffusion module to better propagate the region and boundary-click labels, which helps to enhance the similarity within the region and the discrimination between boundaries. Vast experiments on the popular GrabCut, Berkeley, DAVIS, MS COCO and SBD datasets verified the effectiveness of the proposed method.

Portable Joint Attention Skill Training Platform for Children With Autism

Children with autism are characterized by milestones in joint attention (JA) skill. They fail to understand the directional cue issued by a partner (during social communication), which often results in them reciprocating inappropriately and not completing the JA bid successfully. The directional cues can be gaze-pointing, finger-pointing, etc., toward a target object of interest. Here, in this article, we present a TABlet-based Joint Attention Task (TABJAT) platform that can present JA tasks on a tablet with virtual characters (serving the role of JA administrator), delivering prompting cues (e.g., gaze-pointing, finger-pointing, and head orientation) and virtual objects offering sparkling cue. The platform is adaptive to one's JA skill (while autonomously offering JA tasks with gradually decreasing cueing prompt information, i.e., increasing the task difficulty). This can offer an opportunity for a child to interact with the JA tasks by himself/herself (preserving the triadic interaction regime of JA tasks) and learn the skill while being at home and/or school. Results of a study with 18 children with autism indicate the feasibility of TABJAT to be accepted by the target group, and quantify their JA skill in an individualized manner in terms of their task performance while adaptively offering the JA tasks of varying difficulty.

Automatic Inside Point Localization with Deep Reinforcement Learning for Interactive Object Segmentation.

In the task of interactive image segmentation, the Inside-Outside Guidance (IOG) algorithm has demonstrated superior segmentation performance leveraging Inside-Outside Guidance information. Nevertheless, we observe that the inconsistent input between training and testing when selecting the inside point will result in significant performance degradation. In this paper, a deep reinforcement learning framework, named Inside Point Localization Network (IPL-Net), is proposed to infer the suitable position for the inside point to help the IOG algorithm. Concretely, when a user first clicks two outside points at the symmetrical corner locations of the target object, our proposed system automatically generates the sequence of movement to localize the inside point. We then perform the IOG interactive segmentation method for precisely segmenting the target object of interest. The inside point localization problem is difficult to define as a supervised learning framework because it is expensive to collect image and their corresponding inside points. Therefore, we formulate this problem as Markov Decision Process (MDP) and then optimize it with Dueling Double Deep Q-Network (D3QN). We train our network on the PASCAL dataset and demonstrate that the network achieves excellent performance.

PD40-03 UTILIZATION OF MEDMEASURE!, A SMARTPHONE-BASED APP, AS A RELIABLE METHOD FOR FLUOROSCOPIC AND ENDOSCOPIC MEASUREMENTS

You have accessJournal of UrologySurgical Technology & Simulation: Instrumentation & Technology III1 Apr 2018PD40-03 UTILIZATION OF MEDMEASURE!, A SMARTPHONE-BASED APP, AS A RELIABLE METHOD FOR FLUOROSCOPIC AND ENDOSCOPIC MEASUREMENTS Samuel Washington, III and Maurice Garcia Samuel Washington, IIISamuel Washington, III More articles by this author and Maurice GarciaMaurice Garcia More articles by this author View All Author Informationhttps://doi.org/10.1016/j.juro.2018.02.1928AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES Endoscopic procedures are ubiquitous amongst urology practices across the world. A reliable evaluation of the size of the stone or tumor evaluated during an endoscopic procedure can significantly alter management. To our knowledge a standardized, reliable method for measurement in endoscopic cases has yet to be developed. We describe utilization of a smartphone-based application, MedMeasure! for objective measurements during endoscopy. METHODS We conducted an ex-vivo experiment using an Olympus cystoscope with a 30 degree lens to simulate a standard endoscopic procedure. We then visualized a reference item of known diameter (quarter and nickel) and a target object (ruler) in the same field of view. We describe the reference and target objects in several configurations based upon their position to each other and relative distance to the cystoscope. Nomenclature used as follows: A-P, both objects are perpendicular to the cystoscope; Angled, objects parallel to one another and tilted away from scope; Askew, objects at different angles relative to scope; Equidistant, same distance from cystoscope; Offset, objects at unequal distances from scope; Foreground, one object closer to scope than the other. For each configuration a picture was taken of the image on the monitor using an iPhone camera. MedMeasure! was used to then measure the width of the target object. We calculated the difference between the measured and known diameter of the target object in each configuration. We then derived the percentage distortion between measured and known values by dividing the previously calculated differences by the known diameter. Box plot graphs was used to illustrate the inter-observer variation as well as the extent of distortion of the target image. RESULTS The reference objects used were items whose diameters are well-known and standardized, a quarter, 24.3mm, and a nickel, 21.21mm. The target object of interest was a ruler with a width of 4cm. Distortion was the least when both objects were equidistant, either A-P (12-14% distortion) or askew (6-30%) (p>0.05 for both). The distortion increased significantly when offset or at different distances from the scope (p<0.001 for all). CONCLUSIONS Using a smartphone-based app, MedMeasure!, we demonstrate that accurate measurements can be obtained during fluoroscopic and endoscopic procedures using a reference object with the target at the same distance from the cystoscope. Distortion increases significantly when objects are askew or at different distances from the cystoscope. © 2018FiguresReferencesRelatedDetails Volume 199Issue 4SApril 2018Page: e803-e804 Advertisement Copyright & Permissions© 2018MetricsAuthor Information Samuel Washington, III More articles by this author Maurice Garcia More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...

A Sparse Representation Method for &lt;italic&gt;a Priori&lt;/italic&gt; Target Signature Optimization in Hyperspectral Target Detection

Many target detectors commonly utilize a single a priori target spectral signature as an input. However, the detection results are greatly affected by the quality of the a priori target spectral signature because the spectral variability phenomenon is universal and anisotropic in hyperspectral image data. This paper proposes a sparse representation-based method to generate an optimized target spectrum from limited target training samples, which is able to alleviate the impact of spectral variability on hyperspectral target detection. When lacking comprehensive knowledge about the target object of interest, an optimized representative target spectrum should be expected to be reconstructed by the hyperspectral data themselves in a sparse representation manner following the characteristics of the data structure and then be generated by a set of selected candidate pixels that contain the target signal with a varying status. With the optimized a priori target signature, the experimental results of the detection of different characteristics of objects with three different types of hyperspectral images confirm the effectiveness, robustness, and generality performance of the proposed method.