Visual grounding, a task to ground (i.e., localize) natural language in images, essentially requires composite visual reasoning. However, existing methods over-simplify the composite nature of language into a monolithic sentence embedding or a coarse composition of subject-predicateobject triplet. In this paper, we propose to ground natural language in an intuitive, explainable, and composite fashion as it should be. In particular, we develop a novel modular network called Neural Module Tree network (NMTREE) that regularizes the visual grounding along the dependency parsing tree of the sentence, where each node is a neural module that calculates visual attention according to its linguistic feature, and the grounding score is accumulated in a bottom-up direction where as needed. NMTREE disentangles the visual grounding from the composite reasoning, allowing the former to only focus on primitive and easy-to-generalize patterns. To reduce the impact of parsing errors, we train the modules and their assembly endto-end by using the Gumbel-Softmax approximation and its straight-through gradient estimator, accounting for the discrete nature of module assembly. Overall, the proposed NMTREE consistently outperforms the state-of-the-arts on several benchmarks. Qualitative results show explainable grounding score calculation in great detail. * Corresponding Author. a pink umbrella carried by a girl in pink boots (a) Holistic a pink umbrella carried by a girl in pink boots (b) Tripletpink umbrella carried by girl in pink boots ADJ NN VB ADP acl agent pobj prep pobj amod amod ADP NN NN ADJ Sum[by] Comp[in] Comp[carried] Sum[umbrella] Sum[girl] Sum[boots] Single[pink] Single[pink] Single[ROOT] NMTREE Dependency Parsing Tree (c) NMTREE Region Proposals man VB ADP pobj prep dobj amod nsubj NN ADJ NN NN riding VB dobj Dep. label Embedding Word Embedding POS tag Embedding Concat Fc layer Gumbel Softmax Sum[riding]? Comp[riding]? the man with helmet riding a brown horse Single[ROOT] Sum[man] Comp[with] Comp[riding] Sum[horse] Single[brown]
Islet cell replacement is considered as the optimal treatment for type I diabetes. However, the availability of human pancreatic islets for transplantation is limited. Here, we show that human bone marrow-derived mesenchymal stem cells (hMSCs) could be induced to differentiate into functional insulin-producing cells by introduction of the pancreatic duodenal homeobox-1 (PDX-1). Recombinant adenoviral vector was used to deliver PDX-1 gene into hMSCs. After being infected with Ad-PDX-1, hMSCs were successfully induced to differentiate into insulin-secreting cells. The differentiated PDX-1+ hMSCs expressed multiple islet-cell genes including neurogenin3 (Ngn3), insulin, GK, Glut2, and glucagon, produced and released insulin/C-peptide in a weak glucose-regulated manner. After the differentiated PDX-1+ hMSCs were transplanted into STZ-induced diabetic mice, euglycemia can be obtained within 2 weeks and maintained for at least 42 days. These findings validate the hMSCs model system as a potential basis for enrichment of human beta cells or their precursors, and a possible source for cell replacement therapy in diabetes.
Many vision-language tasks can be reduced to the problem of sequence prediction for natural language output. In particular, recent advances in image captioning use deep reinforcement learning (RL) to alleviate the "exposure bias" during training: ground-truth subsequence is exposed in every step prediction, which introduces bias in test when only predicted subsequence is seen. However, existing RL-based image captioning methods only focus on the language policy while not the visual policy (e.g., visual attention), and thus fail to capture the visual context that are crucial for compositional reasoning such as visual relationships (e.g., "man riding horse") and comparisons (e.g., "smaller cat"). To fill the gap, we propose a Context-Aware Visual Policy network (CAVP) for sequence-level image captioning. At every time step, CAVP explicitly accounts for the previous visual attentions as the context, and then decides whether the context is helpful for the current word generation given the current visual attention. Compared against traditional visual attention that only fixes a single image region at every step, CAVP can attend to complex visual compositions over time. The whole image captioning model -CAVP and its subsequent language policy network -can be efficiently optimized end-to-end by using an actor-critic policy gradient method with respect to any caption evaluation metric. We demonstrate the effectiveness of CAVP by state-of-the-art performances on MS-COCO offline split and online server, using various metrics and sensible visualizations of qualitative visual context. The code is available at https://github.com/daqingliu/CAVP
BackgroundCurrently, a constant shortage in the supply of platelets has become an important medical and society challenge, especially in developing country, and the in vitro production of megakaryocytic progenitor cells (MPs) from cord blood could represent an effective platelet substitute. In the present study, our objective was to determine the safety and feasibility of ex vivo generated MPs in patients.Methods and FindingsMPs were produced and characterized from cord blood mononuclear cells under a serum free medium with cytokines. We investigated the feasibility of expansion and infusion of cord blood-derived MPs in 24 patients with advanced hematological malignancyes. The primary end point was the safety and tolerability of the infusion of cord blood-derived MPs. No adverse effects were observed in patients who received ex vivo-generated cells at concentrations of up to a median value of 5.45×106cells/kg of body weight. With one year follow-up, acute and chronic GVHD had not been observed among patients who received MPs infusion, even without ABO blood group and HLA typing matching.ConclusionsThese initial results in patients are very encouraging. They suggest that infusion of cord blood-derived MPs appears safe and feasible for treatment of thrombocytopenia.Trial Registration
www.chictr.org ChiCTR-TCH-09000333.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.