Syed Zain Masood

Spearheading engineering excellence, pioneering research initiatives, and providing strategic support to sales efforts. With a blend of technical expertise and business acumen, orchestrating innovative solutions that drive growth and exceed client expectations.

Leading a team of Computer Vision PhDs in researching, designing and implementing improved methods for object detection, face recognition and tracking. Focused towards improvements using deep learning solutions that are accurate, fast and have a small footprint (work seamlessly on mobile devices). More specifically, we have designed and implemented industry leading deep learning solutions for: * Object detection * Face recognition * Face attributes (age/gender/emotion/pose/landmarks) * Vehicle make/model/color recognition * License plate recognition You can also find our white papers on arXiv: 1. DAGER: Deep Age, Gender and Emotion Recognition Using Convolutional Neural Network. 2. View Independent Vehicle Make, Model and Color Recognition Using Convolutional Neural Network. 3. License Plate Detection and Recognition Using Deeply Learned Convolutional Neural Networks.

Research in the field of Computer Vision and Machine Learning. Experienced in action recognition, face recognition, shadow detection/removal, computational photography, etc. Projects Include: + Exploring the trade-off between action recognition accuracy and observational latency for Human Computer Interaction (HCI) environments. Used the Microsoft Kinect sensor for capturing actions and proposed a learning framework that maintains a healthy balance between accuracy and latency. + Recognizing faces for popular social networking service by leveraging information available from the social network e.g. albums, friendships, up-loaders, etc. A significant improvement is reported by incorporating such attributes for the task of recognizing faces in pictures. + Highlight the importance of localization for recognizing actions. Show weakness in current recognition systems and an in depth analysis of why current techniques suffer. Propose an automated localization probabilistic system which outperforms the state-of-the-art on renown action datasets. + Built a system for detecting shadows in monochromatic images. Used shadow variant and invariant cues from illumination, texture and derivative characteristics with a Boosted Decision Tree (BDT) classifier integrated with a Conditional Random Field (CRF). + Devised and implemented an efficient and effective method of estimating the correct response of saturated pixels in an image. Used a MRF framework for estimating the correct response using non-saturated color channels ratios of a pixel and its neighbors.

Worked as part of the Computer Vision Group. + Involved in developing a better action localization and recognition system for the VIRAT public dataset. + Experimented with a successful localization approach on the VIRAT dataset. Highlighted the high complexity of the dataset and improvements needed in the method for improved results on surveillance datasets like VIRAT public dataset.

A part of the Generator Engineering Computing Group. I was responsible for: + Working with the backend of the Siemens Integrated Design (SID) Environment. + Porting SID methods to multiple platforms (Unix / Linux / MS-Windows). + Testing SID methods on a different compiler (G95) and applied unit conversion (USCS↔SI).

As a member of the Protocol Sequence Re-player Project team. We were responsible for constructing a tool to help detect malicious behavior for LAN protocols including proprietary protocols for Windows, and Linux. I personally: + Worked with the network security applications team on a protocol analyzer tool namely Protocol Sequence Re-player. + Developed and documented the XDMCP, X11, FTP and WinReg protocol plug-ins.