Sayem Mohammad Siam

My Resume

Work Experience

I am Sayem Mohammad Siam, currently working as a software engineer at Google LLC, USA. Previously, I also worked at Meta Inc. also known as Facebook. I also worked as a machine learning software engineer at Huawei Technologies, Canada. I have completed my master's degree in computer vision and robotics from the department of Computing Science, at University of Alberta.
- 2022-Current
  
  Software Engineer / Google LLC
- 2021-2022
  
  Machine Learning Software Engineer / Meta Inc (AKA Facebook)
- 2018-2021
  
  Machine Learning Software Engineer / Huawei Technologies, Canada
  
  Project
  
  Designed and developed a key video processing module using C++, JNI, OpenCV, Android, Tensolflow, which automatically cuts a video into different segments based on the characters, of video processing lib, released in Huawei flagship phones
  
  Designed and developed a gesture library which uses its camera to recognize mute, play, pause and video navigation by recognizing hand gestures and finally released in Huawei smart-tv (C++, OpenCV, Tensolflow)
  
  Single-handedly developed and delivered an image gallery app, XAlbub, which provides an efficient way of image searching based on persons appear in the images and patented the idea (C++, JNI, OpenCV, Android, Tensorflow)
  
  Implemented, parallelized (OpenCL) and optimized different computer vision algorithm such as image segmentation, image pyramids, object detections and object tracking
  
  Implemented and optimized different deep learning based models and deployed on Android \textbf{(SSD face detector and SiamFC tracker)} \item Implemented a optical flow and Kalman filter-based tracker which improved the \textbf{runtime by 30\%}
  
  Wrote a thread pool module for gesture lib to highly parallelize the system which improved the \textbf{runtime by 50\%}
  
  Wrote a custom json parser \textbf{using C++} which enable to pass hundreds of configuration parameter from Android to C++ library and converts any C++ class to json string and which also improved the code readability and reduced codebase complexity
  
  Improved overall \textbf{runtime by 10\%} by refactoring the existing code and improved code readability
- 2014-Current
  
  Research & Teaching Assistant / University of Alberta
  
  Advance Game theory(CMPUT 350), Introduction to Digital Image processing(CMPUT 206), Introduction to Fundamentals of Computation(CMPUT 175), Introduction to Data Structure(CMPUT 175)
  
  Research Project
  
  Bear Detection Using Convolution Neural Network (Python, Matlab, AWS)
  – Used CNN model (pre-trained on ImageNet dataset) to extract image feature
  – Compared the result between HOG+SVM and CNN+SVM
  
  Gender and age group classification using Markov Random Field (Matlab)
  – Used Gaussian Markov Random (MRF) Fields to build functional connectivity models
  that can be used as classifiers using resting-state fMRI data
  
  Existing algorithms comparison for 3D Mapping and Tracking (Python and ROS)
  – Compared performance of PTAM and DTAM algorithm for different dataset
  
  SLAM using Recursive Bayesian Filter (Python and ROS) git link
  – Used Global image descriptor for image matching score
  – Used image matching score as observation likelihood in Bayes filter and Geometric verification for loop-closure nodes.
  
  Auto-pilot for Husky and Kingfisher Robots (Python and ROS) git link
  – Users can select any path using Google map and robot will follow that path with avoiding obstacles
- 2012-2014
  
  Software Engineer / Samsung Electronics, R&D Institute
  
  Worked in the following projects
  
  Android Benchmark Application(Java, OpenGL)
  
  Android Remote Access Application(Java)
  
  Native Application Development for Tizen OS (www.tizen.org)(C++)
  
  Tizen Gear App(HTML, Javascript, jQuery)
  
  Web Application Development for Tizen OS(html, css, javascript)
  
  Tizen OS HAL layer Testing(C)
  
  Tool development which will automatically test the HAL of Tizen OS for different devices(C++)
  
  Sound Engine Prototype for Tizen OS(C++)
  
  Idea Prototyping: Human fingertip detection, tracking and gesture recognition(Matlab)
  
  Unit test case development in Tizen Native API(C++)
My Education

I am doing my M.Sc. degree in the department of Computing Science at University of Alberta. My master thesis is on Robot localization and merging multiple maps. In my thesis, I propose a new algorithm called Fast-seqSlam which has a time complexity of O(nlog n), faster than the existing algorithms and has the same performance comparing with the existing algorithms
- 2016-Present
  
  M.Sc. in Computer Science / University of Alberta, Edmonton
  
  M.Sc. in Computing Science,Thesis Based, GPA – 3.7/4.0
  
  Coursework
  
  Introduction to Machine Learning, Robotics - Visual Navigation, Probabilistic Graphical Model, Computer Vision, Reinforcement Learning (audit),Teaching and Research Methods
- 2008-2012
  
  B.Sc. Engineering in Computer Science / Islamic University of Technology, Bangladesh
  
  B.Sc. Engineering in Computer Science and Engineering, GPA – 3.85/4.00
- Present
  
  Masters / Thesis
  
  Title: Simultaneous Localization and Mapping for multiple robots
  
  Multi-robot SLAM is very important in Co-operative robotics. Multiple robots can explore the environment more robustly and efficiently than a single robot. To work together they need to build a global map using all the local maps. In this paper, we demonstrate that our algorithm works faster than the existing algorithms for the map merging.
  
  Supervisors: Professor Hong Zhang
  
  Research Interests
  
  Robot Localization, Probabilistic Graphical Model, Supervised and Unsupervised Learning, Reinforcement Learning
Volunteer Activities
- Developed IUT-4th National ICT FEST’s website, using html, css, php and mySql
  
  Arranged programming workshops at Samsung R&D Institute Bangladesh.
  
  Conducted Linux training at Samsung R&D Institute Bangladesh for the new Interns.
  
  Organizing member of IUT-4th National ICT FEST 2012
  
  Organizer of ’Get Root Cyber Security Contest 2012’
- OIC Scholarship for Full Free Undergraduate studies at Islamic University of Technology
  
  Board of Intermediate and Secondary Education Scholarship
About Me
Personal Information

NameSayem Mohammad Siam

Date of birth3 July 1990

Phone1-780-807-7864

Email siam@ualberta.ca

WebMy Homepage

AddressCSC: 3-51, CIMS Lab Department of Computing Science University of Alberta Edmonton, Alberta Canada T6G 2E8
My Skills

java

90%

Python

85%

C&C++

70%

HTML/CSS

85%

PHP

75%

MySQL/SQL Server/Oracle

70%

JavaScript/Jquery

75%

Download CV

Programming Contest

In Samsung Sotong S-topcoder sotong.sec.samsung.net, my last ranking was 43 among 20452 global Samsung contestants
My Research Papers
- Sayem Mohammad Siam and Hong Zhang, "Fast-SeqSLAM: A Fast Appearance Based Place Recognition Algorithm." 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, May 29 - June 3, 2017 [paper]
- Human Computer Interaction Using Marker Based Hand Gesture Recognition [paper]
My Research Projects

Fast-SeqSLAM: A Fast Appearance Based Place Recognition Algorithm

Master's Thesis

Fast-seqSLAM

Loop closure detection or place recognition is a fundamental problem in robot simultaneous localization and mapping (SLAM). SeqSLAM is considered to be one of the most successful algorithms for loop closure detection as it has been demonstrated to be able to handle significant environmental condition changes including those due to illumination, weather, and time of the day. However, SeqSLAM relies heavily on exhaustive sequence matching, a computationally expensive process that prevents the algorithm from being used in dealing with large maps. In this paper, we propose Fast-SeqSLAM, an efficient version of SeqSLAM. Fast-SeqSLAM has a much reduced time complexity without degrading the accuracy, and this is achieved by using an approximate nearest neighbor (ANN) algorithm to match the current image with those in the robot map and extending the idea of SeqSLAM to greedily search a sequence of images that best match with the current sequence. We demonstrate the effectiveness of our Fast-SeqSLAM algorithm in appearance based loop closure detection.. [code]

Read More

Human Computer Interaction Using Marker Based Hand Gesture Recognition

Undergrad Thesis

Marker Based Hand Gesture Recognition

Abstract— Human Computer Interaction (HCI) has been redefined in this era. People want to interact with their devices in such a way that has physical significance in the real world, in other words, they want ergonomic input devices. In this paper, we propose a new method of interaction with computing devices having a consumer grade camera, that uses two colored markers (red and green) worn on tips of the fingers to generate desired hand gestures, and for marker detection and tracking we used template matching with kalman filter. We have implemented all the usual system commands, i.e., cursor movement, right click, left click, double click, going forward and backward, zoom in and out through different hand gestures. Our system can easily recognize these gestures and give corresponding system commands. Our system is suitable for both desktop devices and devices where touch screen is not feasible like large screens or projected screens. [code]

Read More

Bear Detection

Group project for Machine Learning (CMPUT 551).Bear classification is a tasking of predict whether an image contains the bear object or not. The problem is challenging because the large variance of bear color appearance and size, which attracts attentions of researchers from both computer vision and machine learning. In this project, we explored two popular machine learning techniques one is convolution neural network(CNN) and other is Support Vector Maching(SVM). The CNN feature based SVM classifier performs best.

Gender and age group classification

application of Gaussian Markov Random Fields to build functional connectivity models that can be used as classifiers using resting-state fMRI data. These models present at least three advantages: 1) There is a natural mapping from brain data to the graphical model. 2) There are efficient algorithms to train the model. 3) It is possible to directly evaluate model fit of the data. We test our methodology in two binary problems: gender classification (male vs. female) and age group classification (people younger than 10 vs people of at least 17 years old).

3D object Reconstruction and the existing algorithms comparison

Group project for the course Computer vision. Live motion and structure estimation from a single moving video camera has potential applications in domains such as robotics, wearable computing, augmented reality and the automotive sector. In this project, we compared the accuracy and performance of existing real time 3D reconstruction algorithm. There are many existing systems like DTAM, PTAM, Microsoft Photosynth, KinectFusion. We studied different existing application for 3d reconstruction but in our project, we only compared the result of these two applications, PTAM and DTAM.

SLAM using Recursive Bayesian Filter

We used sensor odometry and visual images to build robot poses and constraints and relies on Bayesian filtering to estimate loop closure probability

Auto-pilot

We demonstrated this project in the NCFRN Field trial 2015. We implemented it for Husky and Kingfisher robots. We select the route in Google map then robot follows the map by itself.
My Project Source Code
- Ros package to extract bag files. Topics name are specified for our Husky robot https://github.com/siam1251/rosbag_extract
- Marker Based Hand Gesture Recognition [code] [paper]
Contact Information

Mailing Address:

CSC: 3-51, CIMS Lab
Department of Computing Science
University of Alberta
Edmonton, Alberta Canada
T6G 2E8
Office: 4-18 Athabasca Hall
Email: siam@ualberta.ca
Phone: 1-780-807-7864

My Resume

Work Experience

Software Engineer / Google LLC

Machine Learning Software Engineer / Meta Inc (AKA Facebook)

Machine Learning Software Engineer / Huawei Technologies, Canada

Research & Teaching Assistant / University of Alberta

Software Engineer / Samsung Electronics, R&D Institute

My Education

M.Sc. in Computer Science / University of Alberta, Edmonton

B.Sc. Engineering in Computer Science / Islamic University of Technology, Bangladesh

Masters / Thesis

Volunteer Activities

Awards

About Me

Personal Information

My Skills

Programming Contest

My Research Papers

My Research Projects

Fast-SeqSLAM: A Fast Appearance Based Place Recognition Algorithm

Human Computer Interaction Using Marker Based Hand Gesture Recognition

Bear Detection

Gender and age group classification

3D object Reconstruction and the existing algorithms comparison

SLAM using Recursive Bayesian Filter

Auto-pilot

My Project Source Code

Contact Information

Mailing Address: