Search

In recent years, deep learning techniques (e.g., U-Net, DeepLab) have achieved tremendous success in image segmentation. The performance of these models heavily relies on high-quality ground truth segment labels. Unfortunately, in many real-world problems, ground truth segment labels often have geometric annotation errors due to manual annotation mistakes, GPS errors, or visually interpreting background imagery at a coarse resolution. Such location errors will significantly impact the training performance of existing deep learning algorithms. Existing research on label errors either models ground truth errors in label semantics (assuming label locations to be correct) or models label location errors with simple square patch shifting. These methods cannot fully incorporate the geometric properties of label location errors. To fill the gap, this paper proposes a generic learning framework based on the EM algorithm to update deep learning model parameters and infer hidden true label locations simultaneously. Evaluations on a real-world hydrological dataset in the streamline refinement application show that the proposed framework outperforms baseline methods in classification accuracy (reducing the number of false positives by 67% and reducing the number of false negatives by 55%).

The education directors from fifteen museums and performing companies in Boston and Springfield, Massachusetts, were trained to develop and pilot career education programs for K-12 students using the resources of their institutions. The project was divided into three segments and, during each segment, a series of training seminars were conducted to meet the specific objectives of the segment. Segment one acquainted the education directors with the general assumptions and goals of career education. Each director developed a career information package which incorporated the skills, training requirements, and attitudes which characterized jobs and careers within and related to their respective institutions. The second segment provided the directors with instructional skills to enable them to design and pilot career education programs. This involved joint planning sessions with partner schools. In the last segment, to insure the continued involvement of these cultural institutions in career education, the directors developed long-range plans for the implementation of career education programs at their institutions. A project evaluation indicated that the objectives were all essentially achieved. (This report discusses the training projects carried out during the project, analyzes the career information packages, and generally describes the pilot programs and the long-range plans. A report regarding the training seminars is appended.) (EM)

Software for computer assisted instruction in writing (CAI/writing) can be described in terms of "generations." The earliest (because easiest) application of computers to writing instruction was in the area of drill and practice. By focusing on one or another element of basic skills, well-designed drill and practice software provides instruction that takes advantage of the computer's capacity for presenting information in a compelling and interactive way while keeping track of students' progress. Second generation software--or writer aids--refers to software that focuses on one or another stage of the composing process: prewriting, writing, or rewriting. This generation, in addition to including prewriting programs and using word processors, provides programs that will analyze the surface features of the students' text and give feedback on such dimensions as the degree of sexist language, jargon, or imprecise diction. Third generation software includes integrated author systems. In addition to these more traditional approaches, there is the growing availability of telecommunications networks that provide students with the opportunity to write to others, get responses, and get real responses from real readers. The field of CAI/writing is fortunate in that it has plenty of good models to choose from in the instructional arena. (HOD)