Tasks: Build a face recognition model (from the child’s images) Stream Alternative (extention): upon confirmed detection, theĬloudlets might also provide a tracking functionality of the child. The child might move far from the current location until detection isĬonfirmed. Once a match is detected the parents will be notified Images and use it to match faces on videos and images coming from theĬrowd’s input. The cloudlet will build a classifier from the child’s Parents will have to share aįew recent photos of the missing child with a cloudlet server and askįor help from the crowd around the area where the child is believed toīe missing in.
This project aims to reduce the burden on desparate parents andĪlleviate their unpleasant experience. Students: Sahil Shah, Aditya Gabbita Mentor: Khalid Elgazzarįinding a missing child in the crowd is a crucial and non trivial The goal of the project is to demonstrate some sort ofĪpplication showing an enhanced live view of what the camera is Many algorithms to do this, including some that are supported in Provide clues to the ground truth at a sub-pixel level. Superresolution is that multiple images that are slightly offset can On a mobile phone or Google glass device. Produce a “live” view that is higher resolution than the native camera This project will attempt to apply superresolution techniques to Students: Joel Cao, Caglayan Gemici Mentor: Babu Pillai Pointer on real objects or projected screen. Alternatively, the system can beĭesigned to respond to other camera-sensed inputs, such as a laser The goal of this project is to demonstrate some sort of This will not run in interactive time on a laptop, and will needĬloudlets.
#Coco disk archive code
Vision code used volumetric template matching – a fairly simple, butĬomputationally expensive algorithm that is fairly easy to train. This system used an unmodified tetris-style game,Īnd a pair of cameras to allow two players at a time. Example: Gestris (gesture controlled tetris) fromĪ few years ago. The game using large hand and body motions, observed by a fixed camera This project seeks to use camera based sensing of gestures in an Students: Subramanian Natarajan, Priyanka Kulkarni Mentor: Babu Pillai Camera-based recognition ( Poster, Video) Up downloading of the image data (reference demo video:ĥ. The cloudlet will provide a cache for the Google Street View to speed A user willĬhoose the route interactively as the hyper-lapse video proceeds and This work to enable an interactive hyper-lapse system. A reference prototype is implemented and open sourced at Will implement interactive hyper-lapse simulation using Google Street In addition to the latency benefit, a cloudlet is a perfect place forĬaching large amount of data with locality. Students: (b) Varun Saravagi, Harsha Rastogi ( Poster, Video) Students: (a) Junjue Wang, Xinkai Wang ( Poster, Video) Interactive Google Street View hyper-lapse simulation Skills Learned: Python, Caffe (C++/Python) or Torch (Lua), OpenCVĤ. The streaming can optionally be done with the Gabriel framework: State-of-the-art computer vision models using deep neural networks This project will use existing open source implementations of The Google Glass device will stream video to a cloudlet, which willĭetect objects in the video frames and send object names back to theĭevice. Glass users by providing spoken cues of objects they are looking at.
This project provides naive assistance to visually impaired Google Students: Shruti Dhoot, Vrushali Bhutada Mentor: Brandon Amos
Blind person assistance using object recognition ( Poster, Video) Rewritten if we use non-Android based cameras.Ģ. System can be built on top of Gabriel, but the client part has to be Students may also develop new algorithms (e.g. Similar to the “Camera-based recognition” project below, this projectĬould use existing implementation on volumetric template matching. Is to analyze a user’s action and give user real-time feedback. It could alsoĬount and record different activities a user has performed, The goal This project aims at using environmental cameras to detect a user’sĪctivities inside a gym, e.g. Students: Misha Kutsovsky, Sakshi Goel Mentor: Zhuo Chen Environmental Camera Based Virtual Gym Coach ( Poster, Video) Posters and YouTube videos of final demos from previous classes
0 kommentar(er)
