Want to join CVI Group? Get in touch with us.
The Computer Vision and Intelligence Group of IIT Madras was started on September 2009 with a vision of building a
team of students with deep expertise in the technology of Computer Vision. The idea for the formation of such a club was seeded in 2008 when the IIT Madras team had represented India at the International Aerial Robotics Competition (IARC). CVIG which competed against elite teams, from other top Universities in IARC 2009, was acknowledged as the best vision team among all the participating teams.
Being the only club out of which a start-up has grown, CVIG has extraordinary mentorship and motivated and committed members, who have completed Industrial Projects by ITC, The Indian Railways, VDime, Eye hospital CHECK and multiple machine learning projects.
- Fun with Faces – This project was a small scale attempt at Snapchat with a few filters which would add certain add ons to your face. The final version used a library called dlib, which provides facial feature points and using a few machine learning algorithms, real time results were achieved.Initial prototype –
- Rubik’s cube solver – Using images of the sides of a Rubik cube, this project will be able to solve it within seconds and it will give a 3d simulation and the steps required to solve it.
- Expression Morphing – This project could morph expressions to one’s face, to make him smile or frown. It used a few libraries which would help obtain the facial keypoints and after a certain lines of code provided a final image which would have an expression decided by the user.
- Cell Boundary Identification – This project, was a part of Government of India’s innovation challenge. It aims at automated identification of cell boundaries from the pathological slides. This holds a tremendous potential for cancer diagnosis. It involves decomposing a video of the sample into frames and identifying cell boundaries.Input image:
- Sub-pixel Super resolution – This project was aimed at implementing a CSI (TV Serial) style Resolution enhancer for images. A Deep Learning Super Resolution approach was adopted, and traditional transposed convolutions, generally used for upsampling in Deep Learning, was ditched for an Efficient Subsampling Method based on the Phase Shift approach.
- Integrated detection and tracking of multiple objects – Object Detection and localisation tasks on images have posed a lot of challenge to Computer Vision Scientists for a long time. With the advent of Artificial Neural Networks, there was an efficient solution for these tasks. In this project we have attempted to use the state-of-the-art Convolutional Neural Networks to detect objects and localise them in certain frames of videos, while using conventional methods of CamShift and MeanShift from OpenCV to track them in the rest of the frames. This combination ensures good accuracy and enables good real-time speed up. In an attempt to achieve good computational efficiency while maintaining good accuracy, we present this project.
- Face Liveliness detection – Face recognition systems are widely used at a commercial scale for security accesses. Though the systems have grown quite good at the recognition task, they are quite prone to intentional hacks. We attempt to build a face Liveliness system that distinguishes between a real face and a copy of a face.
- Automatic Waste segregation – To design an intelligent dustbin which would segregate waste into different categories with the help of a camera and other sensors, using deep learning.
- Hand Gesture recognition for swarm bots – With the ever growing popularity of usage of drones for consumers and businesses, the inevitable necessity of fleets/swarms of drones being deployed, is in the very near future. Swarm technology is an approach to coordination of bots within a multi bot system. It was inspired by studying ants and other tiny organisms to know their behavior and apply the same to make multiple robots coordinate to perform a certain task.The team is trying to use image processing (Open CV) for recognising the drones indoors.They are also working on various communication protocols for coordination among the various drones.
- Face Recognition – This project is a part of CFI Jarvis which aims at making the new CFI building smart. This project focuses on the security features of the new building. It will continuously scan the environment and grants access to those who have been registered in the database of users. Future features will include face search through the institute database and alerts in the case of an attempt to breach security protocols.
- Voice App Control – This project, uses one’s voice to control mobile phone apps. By using Natural Language Processing techniques, this app will be able to seamlessly control your phone’s interface.
- AutoCaption – This project, as the name suggests, automatically provides captions to an input image using state of the art neural networks. It also provides hashtags.
- Video Stabilization for UAVs – The primary objective of the project is to ensure proper clarity of a video taken from a moving camera source. This will help in retrieving useful features which otherwise will be rendered useless due to camera shake.
- Evaluation of OMR – Using computer vision and machine learning techniques, one could make the process of evaluating the OMR easier and more accurate compared to the traditional methods that currently exist.