Pankaj Gupta's repositories
Cloud-Game
Cloud Game is an augmented reality game that uses Computer Vision along with some mechanism for interaction for having 3D interaction in augmented scene. In this report I present a 3D game involving human body, its pose and gesture recognition, a basic skeleton representation of human body to assist pose and gesture recognition and to represent it as a player while interacting with the augmented 3D scene. The game is a single player 3D game where in the player have to first pose and follow certain body gesture in a sequence to obtain a weapon (a sword in this case) in his/her hand and then use this weapon to burst/break the objects (a beach ball in this case) flying past in the augmented scene (in right to left direction with respect to actual player) to earn rewards. Un-burst objects crossing off the view will again be spawned at their starting position and will again fly right to left direction holding lesser reward. Objects follow random directions in the 3D space and hence player would have to move in 3D space to burst the object after which the object would disappear from the scene.
OpenGL-Painter
OpenGL-Painter is a Augmented Reality game. It is a real-time vision system that uses camera and Wii Remote (referred as Wiimote henceforth) as input to a game-play. Assignment mandates use of a camera and an inertial input device (like Wiimote). Instead of using an existing game (and plugging-in the inputs from my vision system) I preferred writing my own game because I wanted to handle the interaction in game-play. Vision system takes live stream from camera, orientation and button click from Wiimote. Camera live stream is handled using Opensource OpenCV library to detect a human face using Haar classifier cascade features and the live stream forms the background of the game screen. Orientation and button click input from Wiimote is handled using Opensource WiiYourself! Library and represents orientation of a play-tool ( a cannon in the game) and fire event of the cannon respectively. The game is a two player 3D game ( 2D was minimum required as per assignment requirements) with one player(A) using Wiimote to fire cannon on the face of other player(B). So the game represents “Fire and Dodge” behavior. Player A is supposed to fire the cannon on face of player B and aim of player B is to avoid the cannon falling on his/her face. 3D perspective of the game view changes according to the head position of player B and a red flash appears when the cannon fire hits the face.
pankajkgupta.github.io
Github Pages template for academic personal websites, forked from mmistakes/minimal-mistakes
Real_Time_Head_Pose_Estimation_from_Consumer_Depth_Cameras_Client_Server
Modification of code at http://www.vision.ee.ethz.ch/~gfanelli/head_pose/head_forest.html#code to make it Client Server based. Aim is to let multiple Kinects send their data to one server and let server calibrate the feature
road-sign-detection
A program to detect road signs in images using opencv
CaImAn
Computational toolbox for large scale Calcium Imaging Analysis, including movie handling, motion correction, source extraction, spike deconvolution and result visualization.
flask-video-streaming
Supporting code for my article on video streaming with Flask.
kinect_skeleton_dump
Read skeleton information from KinectV2 live stream and dump into csv file along with relative timestamp
kinetics-i3d
Convolutional neural network model for video classification trained on the Kinetics dataset.
pycapture_multivideo
Record multiple ptgrey firewire cameras using PyCapture python library. Send sync trigger on parallel port.
pykinect_record
Record kinect data using PyKinect
RFID_Reader
Simple RFID tag reader code that runs in a loop and appends in a tag file
serving
A flexible, high-performance serving system for machine learning models
tensorflow
Computation using data flow graphs for scalable machine learning
tf-pose-estimation
Deep Pose Estimation implemented using Tensorflow with Custom Architecture for fast inference.
ubc_neuro_faculty_database
UBC Neuroscience faculty databse
videofig
Lightweight image sequence visualization utility based on matplotlib
VideoPlayer
Python/OpenCV based Video Player with keyboard and mouse controls. Useful for Research purposes when one wishes to monitor the video frame by frame.