Explore Projects

Pharo provides an immersive and interactive development environment with powerful refactoring tools. However, its current refactoring workflows can be fragmented, rely on frequent pop-ups, and offer limited feedback, making them less user-friendly. This project aims to improve the usability and interactivity of Pharo’s refactoring tools, with a focus on rename operations. To achieve this, the project will develop a flexible driver and preview infrastructure that supports safe experimentation, introduce live name editing with immediate validation, and enhance the visualization of changes through a hierarchical tree structure. The expected outcomes include a streamlined refactoring workflow, a clear hierarchical preview of all changes, real-time validation of user input, and a more intuitive and interactive user experience that helps developers better understand and manage code transformations.

<p>Magneto- and Electro-Encephalography (M/EEG) are neuroimaging techniques that non-invasively record human brain activity typically using ~ 100-300 simultaneous sensors sampled at 1kHz. Because these sensors are positioned outside subjects' head, they capture a complex mixture of neural sources, and can therefore be particularly challenging to analyze.</p> <p>Machine learning algorithms recently proved to be extremely efficient at decoding these high-dimensional signals - that is: at inferring the neural and cognitive mechanisms and dynamics underlying the M/EEG recordings.</p> <p>The aim of the present project is to interface the signal processing functions already implemented in the MNE-Python MEG analysis library with the scikit-learn machine-learning library. Specifically, we will implement a series of transformers and a dedicated pipeline with the same Application Program Interface (API) as scikit-learn.</p>

<p>In statistics, linear regression is typically used for modelling relationships between predictor variables and a response variable. In particular, by determining the strength of the relationship between predictors and response variables, linear regression algorithms can explain variation in the response variable, which can be attributed to variation in the predictors, allowing the identification of variables and/or subsets of data that contain relevant information about the response variable. To date, linear regression functionality in the premier toolbox for analysing neural time series in Python, MNE, is capable of handling designs mostly characterised by the introduction of categorical predictors, to explain variation in brain activity, based on ordinary least squares estimation. The goal of the GSoC project is to extend the functionality and inference options of the linear regression module in MNE-Python by providing a set of functions for specifying and testing more complex variants of the linear regression framework, that are commonly used by the neuroscience community. In addition, the project aims at validating these tools on open data resources.</p>

<p>DIPY is an open source software mainly focusing on diffusion Magnetic Resonance Imaging (dMRI) analysis. For nonrigid image registration, it implements a Symmetric Normalization framework. The major goal of this framework is to calculate an affine or deffeomorphic transformation to register two 2D images or 3D volumes. As a well-known problem, registration is a non-convex optimization problem. It is very important to initialize as close to the solution as possible. So in order to avoid getting stuck at local optima, it uses a multi-resolution strategy by building a Gaussian Pyramid. That is to say we can register our images in as many stages as we want, providing previous results as initialization for the next. As this way, we need to do a lot of iterations, thus it slow down the optimization process. To speed up the calculation, one solution is to convert to a parallel computing algorithm. The OpenMP API supports multi-platform shared-memory parallel programming in C/C++ and Fortran. So in this project, I will implement a multi-thread algorithm using OpenMP to improve the performance of nonrigid image registration in DIPY.</p>

This project extends the open-source Cormas simulation platform by integrating a computer vision system that allows real-world interaction with agent-based models. The goal is to enable users to turn Cormas simulations into interactive, game-like experiences by detecting and tracking physical objects (such as game pieces) on a board and linking them to simulation agents. To achieve this, I will develop a lightweight Python-based vision pipeline using YOLO for object detection and tracking, then build a JSON-based interface to communicate with Cormas in real time. On the Pharo/Cormas side, I will implement the necessary APIs and tools to interpret the visual data and control agent behavior accordingly. Deliverables include: - A reusable computer vision module with manual calibration tools. - A JSON-based interface for real-time communication between Python and Pharo. - Integration into Cormas as a plugin or extension. - Documentation and examples for creating interactive physical games with Cormas. - All work will be contributed to the open-source Cormas repository and designed to support future projects involving physical interaction with simulations.

<p>HDMI (High-Definition Multimedia Interface) is a proprietary audio/video interface for transferring uncompressed video data and compressed or uncompressed digital audio data from an HDMI-compliant source device to an HDMI-compliant sink device. The Project deals with the creation of HDMI Source and Sink IP cores in Python Using MyHDL. These IP cores makes it easy for anyone to build a HDMI Transmitter and Receiver using FPGA. Python's power and clarity make MyHDL an ideal solution for high level modeling. The project will also serve as an useful example for the Python MyHDL project to show a non‐trivial example of an IP core.</p> <p>RISC-V is an open ISA freely available for all types of use. The RISC-V ISA has been designed with small, fast, and low-power real-world implementations in mind, but without "over-architecting" for a particular microarchitecture style. The Project deals with the implementation of RISC-V processor in Python using MyHDL. The project demonstrates the advantages of MyHDL and Python in the field of CPU design.</p>

<p>Beamforming is among the most widely used source localization techniques for MEG and EEG data in neuroscience. This project aims at porting an updated version of the five-dimensional beamformer to analyze time-frequency data in source space (Dalal et al., 2008; implemented in MNE-Python as “tf_lcmv”). The update adopts the Hilbert transform to obtain a continuous amplitude and phase time series per frequency band, in contrast to the previous method that utilizes discrete time windows and does not retain phase information. It also adds a weight normalization option to the LCMV beamformer (neural activity index and unit noise gain), which will be available for other LCMV versions as well. Furthermore, statistical testing of the Hilbert beamformer output using the Wilcoxon rank sum test will be implemented to complete the new Hilbert beamformer pipeline. Due to the size of the MEG data and intensiveness of the calculations, the code will be developed with memory efficiency and parallelization in mind. Interactive visualization of source-localized time frequency representations will complete the Hilbert beamformer routine.</p>

<p>Currently, pgmpy deals with only discrete random variables. In many situations, some variables are best modeled as taking values in some continuous space. Examples include variables such as position, velocity etc.</p> <p>The first part of the project creates a module to represent nodes having a continuous domain representation. These nodes would be used in hybrid networks comprising both continuous as well as discrete random variables. The two important features in this part would be -</p> <ul> <li>Representation of User Defined Continuous Random Variables</li> <li>Methods to convert continuous distributions into discrete factors.</li> </ul> <p>The second part of the project will deal with Gaussian distributions. Gaussians are a particularly simple subclass of distributions that make very strong assumptions, such as the exponential decay of the distribution away from its mean, and the linearity of interactions between variables. Gaussians are a surprisingly good approximation for many real world distributions.</p> <p>There will be support for variables comprising the most popular forms of representation in Gaussian distributions -</p> <ul> <li>Linear Gaussian Distribution</li> <li>Joint Gaussian Distribution</li> <li>Canonical Forms</li> </ul>

<p>Realistic hair or fur is essential when creating a plausible virtual world. In feature animation, this is often used to define the signature look of characters; examples include Pixar’s <em>Brave</em> (Iben et al. 2013), and Walt Disney Animation Studios’ <em>Tangled</em> (Sadeghi et al. 2010; also Ward et al. 2010) and <em>Zootopia</em> (Chiang et al. 2016).</p> <p>Currently, Cycles has a working hair shader (<a href="https://docs.blender.org/manual/en/dev/render/cycles/nodes/types/shaders/hair.html" target="_blank">wiki page</a>, <a href="https://git.blender.org/gitweb/gitweb.cgi/cycles.git/blob/HEAD:/src/kernel/closure/bsdf_hair.h" target="_blank">sources</a>), based on Marschner et al. (2003)’s model. Its several assumptions and simplifications make it inaccurate for light colored hair (d’Eon et al. 2011) as well as for most types of fur (Yan et al. 2015). Furthermore, d’Eon et al. (2011) and Khungurn and Marschner (2017) demonstrated it to not be energy conserving.</p> <p>This project intends to upgrade Cycles’ hair shader to the aforementioned <em>Zootopia</em> shader by Chiang et al. (2016), by porting Pharr (2017)’s implementation. Lukas Stockner has made available <a href="https://github.com/lukasstockner/blender/tree/hairshader" target="_blank">a WIP patch </a>, which may also serve as a basis for this work.</p>

CellBox is a model for predicting cell states under drug perturbations, whose code base was originally written in Tensorflow 1. Since Tensorflow 1 has been replaced by Tensorflow 2 and Pytorch, the aim of the project is to recreate CellBox in Pytorch. There are three main components of the model: the data processing pipeline, the model architecture, and the training process, all of which will be sequentially converted into Pytorch. After the conversion is complete, the new Pytorch model is trained using the same training configurations as the Tensorflow model. The following deliverables define the proposal’s success: the trained Pytorch model should reproduce the same results as the current Tensorflow model; the new Pytorch code base should be configured to be able to run on multiple platforms (Linux, Mac, …); and the github should have clear documentations for installing and running the model.

<p>ROS has support for pocketsphinx, but it is very initial stage. So, this project aims to update the ROS package for pocketsphinx with a full-scale support for the recent features and updates. Also, the current package makes use of many external dependencies for its proper functioning. This leads to a lot of complexities in the installation process and in maintaining the codebase. So, another motivation for this project is to get rid of as many of the external dependencies as possible and maintain a proper documentation for the new package. The existing codebase does not have a proper wiki or documentation, neither for the developers, nor for the end-users. And ROS is all about flexible usability. So, this project would also strive to provide a concise, clear and up-to-date documentation of the entire package.</p>

<p>This proposal is for the implementation of Computer-Aided Language Learning from the browser using Pocketsphinx.js. Using Twee/Twine as the question format, users will complete voice-based assignments based on courses which will be integrated with Moodle and Wikiversity. The user's motivation and effort levels will be monitored and a system which monitors these parameter and keep both of them at an optimum level will be integrated with the assignments. Based on these parameters, hints will be given or with held so that the user progresses better. A community feature will be implemented which will facilitate the formation of peer-learning groups (This can also be in the form of a very basic version of an online role-playing game). Amazon Mturk will be used for the collection of transcription data.</p>

<p>This <em>Collect Pronunciation Dictionaries from Wiktionary</em> project aims to expand pronunciation dictionaries in CMUSphinx for new words and multiple languages from Wikimedia Foundation projects. Current Sphinx dictionaries only support limited words and languages, which is difficult to meet the needs of applications, so expanding dictionaries in Sphinx is an urgent need. It's critical to reuse existing pronunciation dictionaries to improve system performance, support new words appeared recently and more languages. A valuable pronunciation source is Wiktionary, which is a multilingual, web-based project to create a free content dictionary of all words in all languages. Although Wiktionary contains pronunciations for many words and multiple languages in a standard format like IPA, it’s not easy to parse those pronunciations from sources in different page formats and convert phonemes in different languages to one common format like CMUBET which can be used by Sphinx. This project will solve those problems and form at least 10 pronunciation dictionaries which will be tested on several ASR benchmarks for Sphinx.</p>

<p>Prism is a probabilistic model checker, by which, the users can model and analyse real time systems. Prism provides its own modelling language. A user can model a system using this language, see a graphical representation of the model and check the behaviour of the model against the optimal strategy (adversary).</p> <p>The aim of the project is to enhance the support for strategies, which will enable more flexible/complex types of strategies/checks which are currently not supported. It will also allow the users to visualize the behaviour of the optimal adversary in a more user-friendly way, and to modify it and re-check with the model. And it will enhance the support for the features like exporting, importing and re-verification of strategies, which are already available in the tool, but not with full functionality.</p>

<p>In Constructive Solid Geometry (CSG) system, the modeling process is usually represented as boolean operations of some basic primitives such as cubes, spheres, etc. This method is often used in solid modeling system and its operations always construct the basic solid modeling pipeline. Besides, the boundary representation (B-rep) is also a method for representing models by using a collection a connected surface elements. So for the purpose of solid modeling, it is commonly impractical to represent these thin solids exactly as bounded volume, thus an available approach is to assign an implicit thickness to each face and then we can consider the thickness as a face property. Simply speaking, when shooting a ray towards the B-rep model, we can get ‘in’ and ‘out’ hit point from a face while the ‘out’ hit point is a small offset from the ‘in’ one with given thickness value along the ray direction.</p>

<p>Systems Biology Markup Language (SBML) is a standard format for the representation and exchange of biological models between computers. The information in an SBML model file is organized as a list of components encapsulated within the model (e.g. compartments, species, parameters, reactions, etc.). However, SBML is difficult to read, comprehend and interpret by humans directly, and tools are required to provide an abstraction layer to interact with SBML objects and the relationships between them. The objective of the project SBML4Humans is to address this issue by providing an interactive and reactive report for SBML models which will allow humans (experts as well as beginners) to easily comprehend the content of a model. The key features introduced in the interactive report are:</p> <p>i) interactive SBML report with navigation between SBML objects</p> <p>ii) search and filter functionality</p> <p>iii) web application with REST API</p> <p>iv) downloadable standalone reports</p> <p>v) support for hierarchical models</p> <p>vi) support for distributions and uncertainties</p> <p>vii) support for COMBINE archives</p> <p>Such an interactive report will thus significantly enhance the process of reading and exchanging SBML models.</p>

<p>CellDesigner is a modeling/simulating software capable of visualizing, modeling, and simulating gene­regulatory and biochemical networks. The key characteristics of CellDesigner are the solidly defined and comprehensive graphical representation of network models and systems biology markup language(SBML) as a model­ describing basis. However, these two characteristics contradict each other. SBML has a standardized definition for layout of network diagram. Before this was standardized, CellDesigner has created its own which is stored in its own way. In the past GSoC project, CellDesigner plug­in that enables to export SBML Layout. To further expand from this software, I will implement a standalone software that converts CellDesigner annotation to SBML Layout and vice versa. It will allow the user to convert CellDesigner annotation into SBML format without CellDesigner. Also the same layout will be held in CellDesigner even when a model is created with another software (i.e. SBML model with Layout). From above, implementing this software will provide interconvertibility between CellDesigner and SBML Layout and remove redundancy for users.</p>

This project aims to optimize the calculation of multimodal embeddings (text, image, audio, video) by exposing REST API endpoints in OpenVINO Model Server. It combines the transformers library for data preprocessing with OpenVINO for efficient inference execution. The project will include a Mediapipe graph for preprocessing and inference, along with a sample application for image search using embeddings stored in a vector database. The goal is to enable flexible, efficient comparison of different data types based on semantic meaning.

<p>The project aims at dealing with the conversion of <strong>CiceroMark to DOCX (OOXML) and vice versa</strong>. In the first portion, we are gonna improve the add-in by improving the existing features and addition of new features which generally includes the ergo expressions and inline blocks of a template. Apart from these, there will be testing of data to ensure that the correct data is entered into the variable fields to avoid any inconsistency. The second portion involves the conversion of the <strong>OOXML to CiceroMark</strong>. In this, improvements will be made to the OOXML transformer present <a href="https://github.com/accordproject/markdown-transform/tree/algoo-ooxml" target="_blank">here</a>. The current transformer works only for very basic OOXML. We need to extend the functionalities of the transformer to support more ciceromark and commonmark classes like lists, optional, inline, ergo formula blocks. In this way, we will ensure more flexibility and portability with the round trip transformation of the smart clause templates.</p>

Multimodal Embedding Explorer (MMEE) is an interactive platform for exploring, cleaning, and validating large-scale multimodal datasets combining images and text. It unifies data inspection, embedding visualization, and quality diagnostics into a single interface, reducing manual effort and tool fragmentation. MMEE allows users to generate and visualize joint embeddings using transformer-based models like CLIP, revealing semantic inconsistencies and annotation noise. Key modules include data ingestion, embedding computation, 2D/3D projection (PCA, UMAP, t-SNE), and interactive validation through image–caption inspection and anomaly detection—empowering faster, more reliable multimodal AI development.

This project extends the Accord Project Template Playground by implementing an end-to-end, sandboxed logic execution pipeline to solve its inability to test interactive smart contract behaviors natively. By utilizing a secure Web Worker architecture, the platform will safely evaluate user-authored TypeScript logic and validate all states and requests against strict Concerto models directly in the browser. The core deliverables include an integrated Monaco-powered logic editor, an interactive contract runner UI for initialization and request handling, dynamic execution results visualization, and comprehensive multi-state contract templates. Together, these features transform the platform from a static authoring tool into a complete, interactive development workspace for smart legal contracts.

This proposal aims to transform SBOannotator from a static, hard-coded tool into a dynamic, intelligent system for annotating SBML models with Systems Biology Ontology terms.After initial discussions with mentor Nantia Leonidou, who provided valuable feedback on my approach, I've developed a comprehensive plan and initial implementation (PR #2 https://github.com/draeger-lab/SBOannotator/pull/2). By implementing real-time SBO term retrieval, integrating multiple enzymatic data sources, and incorporating LLM-based annotation assistance, this project will significantly enhance both the accuracy and usability of SBOannotator while maintaining its core classification strengths. The addition of a standalone desktop GUI with interactive visualizations will make these powerful annotation capabilities accessible to a broader range of systems biology researchers.

<p>The aim is to provide features that will be useful in automatic polygonal healing of a mesh, which BRL-CAD essentially represents as a bag of triangles (bot). Defects in meshes occur when we import from one file format to another file format (here .g being the native format). They might also occur due to imprecise arithmetic, modelling transformations, designers’ fault, programming bugs, etc. Some of these are inevitable. Thus we require some programmatic method to heal meshes which are defective. There are quite a few features already present to heal meshes, but these are not sufficient. Hence, the project. Existing features include: removal of unused and exactly duplicate vertices, dangling edges, decimation of edges, and fusing of equivalent faces. The features that will added are: checking manifold connectivity, removal of gaps, T - joints, overlaps, large holes, and degenerate triangles.</p>