Video Recognition has drawn great research interest and great progress has been made. A suitable frame sampling strategy can improve the accuracy and efficiency of recognition. However, mainstream solutions generally adopt handcrafted frame sampling strategies for recognition. It could degrade the performance, especially in untrimmed videos, due to the variation of frame-level saliency. To this end, we concentrate on improving untrimmed video classification via developing a learning-based frame sampling strategy. We intuitively formulate the frame sampling procedure as multiple parallel Markov decision processes, each of which aims at picking out a frame/clip by gradually adjusting an initial sampling. Then we propose to solve the problems with multi-agent reinforcement learning (MARL). Our MARL framework is composed of a novel RNN-based context-aware observation network which jointly models context information among nearby agents and historical states of a specific agent, a policy network which generates the probability distribution over a predefined action space at each step and a classification network for reward calculation as well as final recognition. Extensive experimental results show that our MARL-based scheme remarkably outperforms hand-crafted strategies with various 2D and 3D baseline methods. Our single RGB model achieves a comparable performance of ActivityNet v1.3 champion submission with multi-modal multi-model fusion and new state-ofthe-art results on YouTube Birds and YouTube Cars.

The purpose of this study is to build and test a support vector machine (SVM) model to predict for the occurrence of lung radiation-induced Grade 2+ pneumonitis. SVM is a sophisticated statistical technique capable of separating the two categories of patients (with/without pneumonitis) using a boundary defined by a complex hypersurface. Despite the complexity, the SVM boundary is only minimally influenced by outliers that are difficult to separate. By contrast, the simple hyperplane boundary computed by the more commonly used and related linear discriminant analysis method is heavily influenced by outliers. Two SVM models were built using data from 219 patients with lung cancer treated using radiotherapy (34 diagnosed with pneumonitis). One model (SVM(all)) selected input features from all dose and non-dose factors. For comparison, the other model (SVM(dose)) selected input features only from lung dose-volume factors. Model predictive ability was evaluated using ten-fold cross-validation and receiver operating characteristics (ROC) analysis. For the model SVM(all), the area under the cross-validated ROC curve was 0.76 (sensitivity/specificity = 74%/75%). Compared to the corresponding SVM(dose) area of 0.71 (sensitivity/specificity = 68%/68%), the predictive ability of SVM(all) was improved, indicating that non-dose features are important contributors to separating patients with and without pneumonitis. Among the input features selected by model SVM(all), the two with highest importance for predicting lung pneumonitis were: (a) generalized equivalent uniform doses close to the mean lung dose, and (b) chemotherapy prior to radiotherapy. The model SVM(all) is publicly available via internet access.

Recently, convolutional neural networks (CNNs) have achieved great improvements in single image dehazing and attained much attention in research. Most existing learning-based dehazing methods are not fully end-to-end, which still follow the traditional dehazing procedure: first estimate the medium transmission and the atmospheric light, then recover the haze-free image based on the atmospheric scattering model. However, in practice, due to lack of priors and constraints, it is hard to precisely estimate these intermediate parameters. Inaccurate estimation further degrades the performance of dehazing, resulting in artifacts, color distortion and insufficient haze removal. To address this, we propose a fully end-to-end Generative Adversarial Networks with Fusion-discriminator (FD-GAN) for image dehazing. With the proposed Fusion-discriminator which takes frequency information as additional priors, our model can generator more natural and realistic dehazed images with less color distortion and fewer artifacts. Moreover, we synthesize a large-scale training dataset including various indoor and outdoor hazy images to boost the performance and we reveal that for learning-based dehazing methods, the performance is strictly influenced by the training data. Experiments have shown that our method reaches state-of-the-art performance on both public synthetic datasets and real-world images with more visually pleasing dehazed results.

No abstract

Unlike other commercial treatment planning systems (TPS) which model the rounded leaf end differently (such as the MLC dosimetric leaf gap (DLG) or rounded leaf‐tip radius), the RayStation TPS (RaySearch Laboratories, Stockholm, Sweden) models transmission through the rounded leaf end of the MLC with a step function, in which the radiation transmission through the leaf end is the square root of the average MLC transmission factor. We report on the optimization of MLC model parameters for the RayStation planning system. This (TPS) models the rounded leaf end of the MLC with the following parameters: leaf‐tip offset, leaf‐tip width, average transmission factor, and tongue and groove. We optimized the MLC model parameters for IMRT in the RayStation v. 4.0 planning system and for a Varian C‐series linac with a 120‐leaf Millennium MLC, and validated the model using measured data. The leaf‐tip offset is the geometric offset due to the rounded leaf‐end design and resulting divergence of the light/radiation field. The offset value is a function of the leaf‐tip position, and tabulated data are available from the vendor. The leaf‐tip width was iteratively evaluated by comparing computed and measured transverse dose profiles of MLC defined fields at dmax in water. In‐water profile comparisons were also used to verify the MLC leaf position (leaf‐tip offset). The average transmission factor and leaf tongue‐and‐groove width were derived iteratively by maximizing the agreement between measurements and RayStation TPS calculations for five clinical IMRT QA plans. Plan verifications were performed by comparing MapCHECK2 measurements and Monte Carlo calculations. The MLC model was validated using five test IMRT cases from the AAPM Task Group 119 report. Absolute gamma analyses (3 mm/3% and 2 mm/2%) were applied. In addition, computed output factors for MLC‐defined small fields (2×2,3×3,4×4,6×6 cm2) of both 6 MV and 18 MV photons were compared to those independently measured by the Imaging and Radiation Oncology Core (IROC), Houston, TX. 6 MV and 18 MV models were both determined to have the same MLC parameters: leaf‐tip offset=0.3 cm,2.5% transmission, and leaf tongue‐and‐groove width=0.05 cm. IMRT QA analysis for five test cases in TG‐119 resulted in a 100% passing rate with 3 mm/3% gamma analysis for 6 MV, and >97.5% for 18 MV. The passing rate was >94.6% for 6 MV and >90.9% for 18 MV when the 2 mm/2% gamma analysis criteria was applied. These results compared favorably with those published in AAPM Task Group 119. The reported MLC model parameters serve as a reference for other users.PACS number(s): 87.55.D, 87.56.nk