Pinned Repositories
A-network-simulator-implementing-entire-protocol-stack
A simulated model which demonstrates the working of the TCP/IP Networking Model. This project shows proper working of the all layers: 1. Physical Layer: The physical layer is responsible for sending and receiving packets from one physical interface to others. 2.Data Link Layer: The data link layer controls the flow of reception, delimitation and transmission of frames and establishes a communication protocol (Media Access Control) between directly connected systems 3.Network layer: Handles the routing and sending of data between different networks. The most important protocols at this layer are IP and ICMP. We have implemented the Routing Information Protocol-(RIP) in the project for shortest path. 4.Transport layer: Provides the means for transmitting data between the two connected parties, as well as controlling the quality of service. The main protocols used here are TCP and UDP. We have set a probability for packet drop in UDP to show that this protocol is unreliable. TCP is a reliable protocol where no packet drop is shown. A server class is made to enable the 3 services we have included HTTP, SSH, SMTP. 5.Application Layer: The application layer is the highest abstraction layer of the TCP/IP model that provides the interfaces and protocols needed by the users. It combines the functionalities of the session layer, the presentation layer and the application layer of the OSI model. We have included two application layer services DHCP and HTTP
Awesome_Domain_Adaptation_and_FL
Counter-top-Corner-Detection
The robot needed to find the extreme edges of the counter-top and the sink. To detect the corners, we used three methods to obtain the coordinates of 8 corners (4-counter top and 4-sink) as shown in figure. The three methods were: 1. Finding corner points using P4P(4 point algorithm) 2. Finding corner points using RANSAC. 3. Finding corner points using colour segmentation. They are explained below. 4.1 Method 1: Finding corner points using P4P The counter top had round colour lables stuck on them in the four outer and inner corners. (See Figure 13 for an example.) The outer points A, B, C , D in figure 13 were segmented based on color and the coordinates were saved. Using 4 Point algorithm, we were able to re project the outer and inner corners of the counter-top with an average re-projection error of 0.489. We can also compute the 3D locations of these points with respect to the camera coordinate frame. 4.2 Method 2: Finding corner points using RANSAC RANSAC divides data into inliers and outliers and yields an estimate of the counter top plane, computed from a minimal set of inliers with greatest support. We Improved this initial estimate with a Least Squares (S) estimation over all inliers (i.e., standard minimization), and then we found the inliers w.r.t the L.S. fit, and re-estimated the plane using L.S. one more time. We used the 3D points (given by the camera) to find the counter plane using this method. We then identified the pixels corresponding to the plane in the corresponding colour image. To find the corners, we used Harris corner detector. Harris’ corner detector takes the differential of the corner score into account with reference to direction directly, instead of using shifting patches for every 45◦ angles, and has been proved to be more accurate in distinguishing between edges and corners. The implementation was done as follows: 1. Compute image intensity gradients in x- and y-directions.
D-MASTER
Welcome to the official repository of D-MASTER: Mask Annealed Transformer for Unsupervised Domain Adaptation in Breast Cancer Detection from Mammograms. This repository hosts the source code, pre-trained model weights, and benchmark dataset RSNA-BSD1K, supporting research in cross-domain breast cancer detection using transformer-based techniques.
Detection-Classification-and-Semantic_Segmentation-of-apples
Research internship Project on Detection and Semantic Segmentation of apples
Encryption-and-Decryption-between-users-in-Morse-using-a-Flashlight.
Morse code has been employed as an assistive technology, helping people with a variety of disabilities to communicate. Morse can be sent by persons with severe motion disabilities, as long as they have some minimal motor control. An original solution to the problem that caretakers have to learn to decode has been an electronic typewriter with the codes written on the keys. Morse code can also be translated by computer and used in a speaking communication aid. An important advantage of Morse code over row column scanning is that once learned, it does not require looking at a display. Also, it appears faster than scanning.
FATE
Welcome to the official repository of arxiv paper FATE: Focal-modulated Attention Encoder for temperature prediction.
HF-Fed
Welcome to the official repository of HF-Fed: Hierarchical based customized Federated Learning Framework for X-Ray Imaging accepted in MICCAI Worshop 2024.
Pose-estimation-using-4-co-planar-points
Traditional PnP algorithms may not perform well when the points selected are co- planar. However, in the robotics challenge, for the problem of identifying the pose of the counter top, the easiest points to identify are the corners of the counter top. Hence, we decided to use a method proposed by Zhang et al. [3]. In this paper they present a pose estimation algorithm based on four co-planar point correspondences. Given four co-planar points and their corresponding image points under a perspective camera, plus the camera’s intrinsic matrix, the camera’s rotation and translation relative to the object plane is determined directly. In essence, the pose estimation problem is converted to the calculation of a planar homography between the object plane and the image plane. Experiments with both synthetic data and real images verify the correctness of this algorithm. The authors do not provide a downloadable implementation. Hence we had to do implement by ourselves and test it.
TransFed
Welcome to the official repository of TransFed: A way to epitomize Focal Modulation using Transformer-based Federated Learning accepted in WACV 2024.
Tajamul21's Repositories
Tajamul21/Counter-top-Corner-Detection
The robot needed to find the extreme edges of the counter-top and the sink. To detect the corners, we used three methods to obtain the coordinates of 8 corners (4-counter top and 4-sink) as shown in figure. The three methods were: 1. Finding corner points using P4P(4 point algorithm) 2. Finding corner points using RANSAC. 3. Finding corner points using colour segmentation. They are explained below. 4.1 Method 1: Finding corner points using P4P The counter top had round colour lables stuck on them in the four outer and inner corners. (See Figure 13 for an example.) The outer points A, B, C , D in figure 13 were segmented based on color and the coordinates were saved. Using 4 Point algorithm, we were able to re project the outer and inner corners of the counter-top with an average re-projection error of 0.489. We can also compute the 3D locations of these points with respect to the camera coordinate frame. 4.2 Method 2: Finding corner points using RANSAC RANSAC divides data into inliers and outliers and yields an estimate of the counter top plane, computed from a minimal set of inliers with greatest support. We Improved this initial estimate with a Least Squares (S) estimation over all inliers (i.e., standard minimization), and then we found the inliers w.r.t the L.S. fit, and re-estimated the plane using L.S. one more time. We used the 3D points (given by the camera) to find the counter plane using this method. We then identified the pixels corresponding to the plane in the corresponding colour image. To find the corners, we used Harris corner detector. Harris’ corner detector takes the differential of the corner score into account with reference to direction directly, instead of using shifting patches for every 45◦ angles, and has been proved to be more accurate in distinguishing between edges and corners. The implementation was done as follows: 1. Compute image intensity gradients in x- and y-directions.
Tajamul21/Pose-estimation-using-4-co-planar-points
Traditional PnP algorithms may not perform well when the points selected are co- planar. However, in the robotics challenge, for the problem of identifying the pose of the counter top, the easiest points to identify are the corners of the counter top. Hence, we decided to use a method proposed by Zhang et al. [3]. In this paper they present a pose estimation algorithm based on four co-planar point correspondences. Given four co-planar points and their corresponding image points under a perspective camera, plus the camera’s intrinsic matrix, the camera’s rotation and translation relative to the object plane is determined directly. In essence, the pose estimation problem is converted to the calculation of a planar homography between the object plane and the image plane. Experiments with both synthetic data and real images verify the correctness of this algorithm. The authors do not provide a downloadable implementation. Hence we had to do implement by ourselves and test it.
Tajamul21/D-MASTER
Welcome to the official repository of D-MASTER: Mask Annealed Transformer for Unsupervised Domain Adaptation in Breast Cancer Detection from Mammograms. This repository hosts the source code, pre-trained model weights, and benchmark dataset RSNA-BSD1K, supporting research in cross-domain breast cancer detection using transformer-based techniques.
Tajamul21/A-network-simulator-implementing-entire-protocol-stack
A simulated model which demonstrates the working of the TCP/IP Networking Model. This project shows proper working of the all layers: 1. Physical Layer: The physical layer is responsible for sending and receiving packets from one physical interface to others. 2.Data Link Layer: The data link layer controls the flow of reception, delimitation and transmission of frames and establishes a communication protocol (Media Access Control) between directly connected systems 3.Network layer: Handles the routing and sending of data between different networks. The most important protocols at this layer are IP and ICMP. We have implemented the Routing Information Protocol-(RIP) in the project for shortest path. 4.Transport layer: Provides the means for transmitting data between the two connected parties, as well as controlling the quality of service. The main protocols used here are TCP and UDP. We have set a probability for packet drop in UDP to show that this protocol is unreliable. TCP is a reliable protocol where no packet drop is shown. A server class is made to enable the 3 services we have included HTTP, SSH, SMTP. 5.Application Layer: The application layer is the highest abstraction layer of the TCP/IP model that provides the interfaces and protocols needed by the users. It combines the functionalities of the session layer, the presentation layer and the application layer of the OSI model. We have included two application layer services DHCP and HTTP
Tajamul21/FATE
Welcome to the official repository of arxiv paper FATE: Focal-modulated Attention Encoder for temperature prediction.
Tajamul21/Detection-Classification-and-Semantic_Segmentation-of-apples
Research internship Project on Detection and Semantic Segmentation of apples
Tajamul21/Encryption-and-Decryption-between-users-in-Morse-using-a-Flashlight.
Morse code has been employed as an assistive technology, helping people with a variety of disabilities to communicate. Morse can be sent by persons with severe motion disabilities, as long as they have some minimal motor control. An original solution to the problem that caretakers have to learn to decode has been an electronic typewriter with the codes written on the keys. Morse code can also be translated by computer and used in a speaking communication aid. An important advantage of Morse code over row column scanning is that once learned, it does not require looking at a display. Also, it appears faster than scanning.
Tajamul21/HF-Fed
Welcome to the official repository of HF-Fed: Hierarchical based customized Federated Learning Framework for X-Ray Imaging accepted in MICCAI Worshop 2024.
Tajamul21/TransFed
Welcome to the official repository of TransFed: A way to epitomize Focal Modulation using Transformer-based Federated Learning accepted in WACV 2024.
Tajamul21/-Image-Inpainting
Tajamul21/ARM-Cross-Compilation-and-Tracing-using-Qemu-Tool
ARM cross compiler Cross compilation is the process of compiling code for one computer system also known as the target on a different system, called the host. A cross compiler is a type of compiler, that generates machine code targeted to run on a system different than the one generating it. Like we used arm cross compiler to convert from X86 processor to ARM. The Process of creating executable code for different machines is called retargeting. The cross compiler is also called as retargetable compiler. We have used GNU GCC cross compiler. A cross compiler is a compiler capable of creating executable code for a platform other than the one on which the compiler is running. For example, a compiler that runs on a X86 processor but generates code that runs on ARM Processor is a cross compiler. A cross compiler is necessary to compile code for multiple platforms from one development host. Direct compilation on the target platform might be infeasible, for example on X86 using .o file , because those systems contain no operating system. In generalization, one computer runs multiple operating systems and a cross compiler could generate an executable for each of them from one main source. Cross compilers are distinct from source-to-source compilers. A cross compiler is for cross-platform software development of machine code, while a source-to-source compiler translates from one programming language to another in text code. Both are programming tools. The GNU Arm Embedded Toolchain is a ready-to-use, open-source suite of tools for C(As in our case), C++ and assembly programming. The GNU Arm Embedded Toolchain targets the 32-bit Arm Cortex-A, Arm Cortex-M, and Arm Cortex-R processor families.
Tajamul21/Assembler-which-converts-all-37-rv32I-instructons-in-hex-and-binary-machine-code
Tajamul21/Awesome_Domain_Adaptation_and_FL
Tajamul21/CapillaryFlowTracker-Salt-Concentration-Analysis-with-Computer-Vision
Analyze fluid motion and capillary movement with varying salt concentrations using computer vision. This repository contains tools and scripts for tracking and visualizing capillary behavior under different salt conditions, facilitating research in fluid dynamics and salt-dependent phenomena.
Tajamul21/Detecting-trash-on-the-floor
The robot needed to detect trash on the floor and pick it up. The trash consisted of items such as crumpled tissue paper, paper cups, and cans. However, it was assumed that the type of trash was known and it was identified by features such as color, texture, shape, and hardness. The robot needed to apply different techniques for picking up the trash depending on the type of trash. This task was limited to detecting different types of trash on the floor and also detecting the dustbin.
Tajamul21/FaceSkyShadow-Saliency-Enhancement-DigitalImageAnalysis
Image enhancement in python for face, sky, and shadowed saliency.
Tajamul21/Image-Morphing-Image-Pyramid-and-Image-Stylization
Tajamul21/Markings-on-the-Floor-Detection
The robot had a task to clean the markings on the floor. For this task, we learnt how to apply skimage functions for thresholding to an image. Thresholding is a type of image segmentation, where we change the pixels of an image to make the image easier to analyze. In thresholding, we convert an image from colour or grayscale into a binary image, i.e., one that is simply black and white. Most frequently, we use thresholding as a way to select areas of interest of an image, while ignoring the parts we are not concerned with. We used the masks returned by these functions to select the parts of an image we was interested in. We tried various thresholding techniques to segment the marker Figure 15 shows the results obtained for various methods. Some other inferences are shown in Figure 16 and Figure 17. We took the images and performed various threshold techniques to segment the markers. One of the famous is Otsu thresholding. In Otsu thresholding, a value of the threshold is not chosen but is determined automatically. A bimodal image (two distinct image values) is considered. The histogram generated contains two peaks. So, a generic condition would be to choose a threshold value that lies in the middle of both the histogram peak values. We used the traditional cv2.threshold function and used cv2.THRESH OTSU as an extra flag. Figure 18 shows the results obtained by the Otsu method. To estimate the depth of the centriod, we used depth data from the the RealSense camera.
Tajamul21/Old_Portfolio_Website
Tajamul21/PoseWatch-Advancing-Real-Time-Human-Pose-Tracking-and-Juxtaposition-with-Deep-Learning
Human pose estimation is the process of continuously monitoring a person's action and movement to track and monitor the activity of a person or an object. Human pose estimation is usually done by capturing the key points which describe the pose of a person.
Tajamul21/Ralith-Milith
Tajamul21/Raspberry-Pi-Object-Detection-autunomous-car
This repository contains code and instructions to configure the necessary hardware and software for running autonomous driving object detection on the Raspberry Pi 4! Details of Software and Neural Network Model for Object Detection: Language: Python Framework: TensorFlow Lite Network: SSD MobileNet-V2 Training Dataset:Berkely Deep Drive (BBD100K)
Tajamul21/reviewing-a-pull-request
Tajamul21/Robot-localization-and-Mapping-using-Particle-Swarm-optimization-PSO-
Tajamul21/www.tajamulashraf.com
Tajamul21/Kashmiri_language_Translation
The Kashmiri Language Translation project focuses on developing a high-quality machine translation system for Kashmiri. We are curating a diverse dataset and using advanced machine learning techniques to improve translation accuracy. The project aims to create a scalable system benefiting the Kashmiri-speaking community.