Link
Github Code
Taking instantaneous decisions during an uncertain situation is near impossible for machines even when considering multi-level observations and their striving ability to learn complex policies. This progress is facilitated by the availability of abundant data, simulators such as games. These projects try to avoid the usage of motors in aerial vehicles(UAV) for motion, instead completely utilize the rising air for fueling motion like bird species. They want to develop a Reinforcement learning agent who can make precise decisions about futuristic atmospheric conditions and numerous other decision making attributes. The UAV will be engineered with lightweight resources along with the agent who deals with conditions, non-stationarity, and learn literally on the fly, with mistakes potentially resulting in drone loss.
Link
This is a proof-of-concept prototype to transform any room to AR experience. This provides gamers a fascinating experience.
RoomAlive presents a unified, scalable approach for interactive projection mapping that dynamically adapts content to any room. RoomAlive brings an exotic AR contents inside a room with flexibility in its performane.
Link
Developing agents who can handle Natural language processing for document summarization, machine translation, question generation, sentence generation, image, and video captioning tasks. Now frameworks for solving NLP is centered at Natural communication, Human interaction through voice and Natural language Understanding. Recent NLP solutions consider language context with Attention mechanisms to obtain comprehensive language understanding. From 2016, DCPs Deep Communication Policies based learning are used to solve a variety of coordination problems, logical puzzles etc. We, therefore, seek to build a DCP based encoder-decoder system for generating longer sequences that are coherent and relevant. We, therefore, seek to build a DCP based encoder-decoder system for generating longer sequences that are coherent and relevant.
Link
As the name suggests, micro-sizing people’s task and achieve micro goals to maximize productivity is the aim of this project. This is an unusual domain, But Chief Scientist Jaime Teevan and Senior Researcher Shamsi Iqbal are working on to negate the idea of devoting a large portion of time to solve huge task instead focus on micro-task accomplishment which in turn contribute to larger productivity.
Link
When similar patients are diagnozied with analgous diseases, doctors have to follow the same treatment. Since all the data are available as Medical records , this project extracts Biomedical NLP information and make it accessible to public. Hence two patients with similar symptoms can be treated similarly with limited doctor’s entailement.
Thus, our first goal is to build systems that can read natural language text to extract biomedical facts, finding the latest research on drug-protein interactions and combing through electronic health records to identify lifestyle and environmental factors. Our research directions include advanced techniques for information extraction, such as deep neural networks that take graph structured inputs. Equally important, though, is to make the knowledge accessible to people. To that end, we are building interfaces to browse and curate the resulting knowledge bases.