Muhammed El-Yamani

I am a

Muhammed

I am interested in projects related to robotics and Artificial Intelligence. I started working in computer vision during my second semester in 2014.
I have a solid background in mathematics and good software skills, in addition to my experience in computer vision, machine learning, and robotics.

I graduated as a mechatronics engineer with a good background in mechanics, control, electronics, embedded system, and other engineer fields.

  • Zamalek, Cairo, Egypt.
  • (+20)106-82-05-697
  • elyamani.business.eng@gmail.com
Me

My Professional Skills

Mathematics, computer vision, Machine learning, Software Skills, Pytorch

Mathematics
Python
C/C++
Software skills
Embedded Linux and command lines
Computer vision
Machine learning
Data skills
Pytorch
Html / CSS/ Java script
Other skills
  • Ros-project




    https://github.com/WikiGenius/ros_project


  • Image-classification

     


    In 2020

    Image classification for flowers (102 classes of flowers)

    Using Pytorch in deep learning and CUDA

    Here is the code







  • Path-planning

    In 2020
    Path planning using many algorithms:

    1_ breadth-first-search: for unweighted edges

    2_ depth-first-search: It has some advantages over other algorithms in rare cases like storage and completeness if there are infinity children for the node.

    3_Dijkstra's: It has the advantage of finding the shortest path between the start node and all other nodes, not only the goal node

    4_ cheapest-first-search: it has an advantage over Dijkstra's in-memory storage but has uniform contours.

    5_ gready_best_search_algorithm: by using a heuristic, it focuses on the goal so, it is faster than cheapest-first-search
    but has drawbacks if there are obstacles, it doesn't guarantee to find the shortest path.

    6_ finally A*: it combines between efficiency and finding the optimal solution but depends on choosing admissible(for finding an optimal solution) and (consistent for efficiency optional) heuristic


    Thanks for

    professor: Peter Norvig, for his good lectures about problem-solving-technology.

    professor: Sebastian Thrun for his support.

    Have fun!

    note:
    Using a Fibonacci priority queue, it has a better performance
    than any priority queue like binary because of insertion


    https://lnkd.in/dQe2CRC



     






  • Building-Neural-Networks From Scratch

     




    In 2020

    From scratch, building Deep Neural networks using libraries such as NumPy, pandas, Matplotlib only.


    The purpose of this project
    is to implement the magical NN from scratch by applying some math like linear algebra, calculus, and probabilities. And understanding the hyperparameters and implementing various techniques to improve the learning process.

    Inventory:
    1_Customize n hidden layers as you want to make it deep
    2_Customize the number of neurons in each hidden layer
    3_Flexibility Adding bias or not for all layers
    4_Early stopping algorithm
    5_Flexibility adding drop out for the layers
    6_regularization L1 / regularization L2
    7_Flavors gradient descent:
       Batch gradient descent
       Stochastic gradient descent
       Mini batch gradient descent
    8_Implement common activation function: Sigmoid / Softmax / Tanh / Relu / Elu
    9_Add Momentum


    https://lnkd.in/dyfJDMB

    Have fun ^_^







  • Kerna personal robot


    In 2018 

    The robot recognizes the object and recognizes human to interact with.

    Features In this project:

    1_ We made an object recognition.
    2_ We made robust facial recognition.
    3_ We made mapping from 2d camera into 3d world to transfer information from computer vision into the physical robot (arm or mobile robot).
    4_ We made a kinematics robot for 6 DOF arm robot.
    5_ We deploy the software on raspberry pi and AVR microcontroller unit.
    6_ We made the mechanism for the robot.
    7_ We made the electric circuits
     
    https://www.youtube.com/watch?v=jk9X27XQk5k&list=PLVU_mZCkhn-aExBlBMJAuHEcXeJb2NYT8&index=5&pp=sAQB&ab_channel=MuhammedEl-Yamani











  • Robot recognize in 3D & grab objects

    In 2018 
     

    We participated in the maker fair with this robot with the FAB LAB.

    Objectives:

    The robot recognizes an object by RGB camera and calculates the distance to the object to grab

    Features:

    1_ computer vision and image processing to recognize the object
    to be robust and invariant to size, rotation

    2_ get some useful physical geometry from the RGB camera 
    like distance,  and mapping cartesian from the camera to the world cartesian, and calculate the orientation of the object.

    3_kinematics of mobile and arm robot

    4_deploy control software into AVR microcontroller.


    https://www.youtube.com/watch?v=OPGJl5cpCm4&list=PLVU_mZCkhn-aExBlBMJAuHEcXeJb2NYT8&index=9&ab_channel=MuhammedEl-Yamani

  • Face recognition

     


    Robust facial recognition in real-time

    using many algorithms and many papers

    In 2018

    https://www.youtube.com/watch?v=Y0jdRN0FLig&list=PLVU_mZCkhn-aExBlBMJAuHEcXeJb2NYT8&index=7&ab_channel=MuhammedEl-Yamani

    https://www.youtube.com/watch?v=3171of1YpuE&list=PLVU_mZCkhn-aExBlBMJAuHEcXeJb2NYT8&index=1&ab_channel=MuhammedEl-Yamani
  • ADDRESS

    Zamalek, Cairo, Egypt

    EMAIL

    elyamani.business.eng@gmail.com

    MOBILE

    (+20)106-82-05-697