A Novel Multimodal Approach for Video Captioning – We propose to use a convolutional neural network (CNN) architecture to learn the structure of a video sequence by means of a sequence-to-sequence architecture. In our model, we utilize CNNs to learn the spatial relationship between images in a sequence to predict the motion of each pixel. We propose a novel learning scheme which requires the pairwise interactions between images which enables a sequential learning process. To solve the spatial relationship problem, we propose an efficient CNN architecture to learn the spatial relationships between images and the temporal dependency between frames. With a novel convolutional architecture, we propose to learn features with large temporal dependency structures on a single CNN which learns a sparse vector representation of frames. We use this representation to learn the temporal dependency structure for learning the temporal dependency structure and use this structure to train a feature model on the spatial relationship between images. Experiments on multiple benchmark datasets demonstrate the effectiveness of our recurrent-CNN architecture in learning temporal dependency structures for video captioning and a video-embedding model.
We consider the problem of identifying movements from unstructured data, and provide a simple implementation on mobile devices in the domain of robotics. To this end, we provide a real-time interactive platform to users (using an embedded computer) able to perform movement identification in real-time and control robot vehicles in real time while maintaining safety and navigation. Our platform provides users an opportunity to access these skills through the interactive robotic interaction, and is the first such platform for autonomous mobility of robots using real-time interactive control and navigation.
A New Way to Evaluate Metrics: Aesthetic Framework
A Bayesian Learning Approach to Predicting SMO Decompositions
A Novel Multimodal Approach for Video Captioning
Deep Learning, A Measure of Deep Inference, and a Quantitative Algorithm
The Classification, GAN and Supervised Learning of Movement Recognition SystemsWe consider the problem of identifying movements from unstructured data, and provide a simple implementation on mobile devices in the domain of robotics. To this end, we provide a real-time interactive platform to users (using an embedded computer) able to perform movement identification in real-time and control robot vehicles in real time while maintaining safety and navigation. Our platform provides users an opportunity to access these skills through the interactive robotic interaction, and is the first such platform for autonomous mobility of robots using real-time interactive control and navigation.