Coding & Development

img2pose is an open-source PyTorch implementation for real-time, six degrees of freedom (6DoF), 3D face pose estimation. This tool uniquely performs face alignment and detection without requiring preliminary face detection or facial landmark localization, simplifying the process. It leverages a Faster R-CNN-based model to regress 6DoF pose for all faces in a photo, even tiny ones. The system allows for visualization of detections, customization of projected bounding boxes, and cropping/aligning faces for further processing. Accepted at CVPR 2021, img2pose outperforms state-of-the-art face pose estimators and even surpasses comparable models on the WIDER FACE detection benchmark, despite not being optimized for bounding box labels.

Gemini vs GPT vs Claude is a dedicated AI comparison tool designed for evaluating the performance of leading large language models. Users can input custom prompts and observe the responses generated by Gemini Pro, GPT-4, and Claude 3. This side-by-side comparison facilitates a detailed analysis of each model's strengths, weaknesses, and unique characteristics, helping users understand their respective capabilities and limitations for various tasks.

ML-GCN is a PyTorch implementation of Multi-Label Image Recognition with Graph Convolutional Networks, as presented in a CVPR 2019 paper. This open-source project provides researchers and developers with the code and pre-trained models necessary to apply GCNs to multi-label image recognition tasks. The implementation highlights improvements achieved by replacing Global Average Pooling (GAP) with Global Max Pooling (GMP) for feature aggregation, demonstrating enhanced performance on datasets like COCO, NUS-WIDE, and VOC2007. It includes detailed instructions for setting up requirements, downloading models, and running demos for VOC 2007 and COCO 2014 datasets, making it a valuable resource for academic research and practical application in computer vision.

Moonlight TV is a community-driven, open-source client for NVIDIA GameStream, specifically designed to bring PC gaming to large screens like LG webOS TVs and embedded devices such as the Raspberry Pi. This lightweight client offers high-performance streaming, ensuring a smooth gaming experience. Its user interface is optimized for large displays and remote controller navigation, making it accessible and enjoyable from the couch. A key feature is its support for up to four controllers, facilitating multi-player gaming sessions. The project emphasizes portability, with successful implementations on macOS, Arch, Debian, Raspbian, and Windows, highlighting its versatility and ease of adaptation to various operating systems. Users can easily install it on webOS via dev-manager-desktop or by downloading IPK/DEB packages from the latest releases.

The POCO C++ Libraries are a comprehensive collection of C++ class libraries designed for building network- and internet-based applications. Conceptually similar to Java Class Library or .NET Framework, POCO focuses on providing solutions to frequently encountered practical problems in software development. It is written in efficient, modern, 100% ANSI/ISO Standard C++ and complements the C++ Standard Library/STL. The libraries are highly portable, supporting a wide range of platforms from embedded systems to servers. POCO is open source, licensed under the Boost Software License, and offers features for JSON, SQL, networking, XML, logging, configuration, HTTP clients/servers, Redis, and MongoDB client access.

RaDe-GS, or Rasterizing Depth in Gaussian Splatting, is a cutting-edge Content & Design tool developed by HKUST-SAIL. It significantly enhances the performance and accuracy of 3D scene reconstruction and rendering by incorporating advanced techniques like multi-view regularization and refined densification strategies. The project provides updated code and formulations, enabling users to achieve superior results on challenging datasets such as DTU and Tanks and Temples. It also supports novel view synthesis and geometry evaluation, making it a powerful resource for researchers and developers working with 3D Gaussian Splatting. The tool is built upon the original 3D Gaussian Splatting implementation and integrates ideas from several recent works to offer a robust and efficient solution for 3D graphics tasks.

DevVerse is a technology solutions provider that previously specialized in AI-integrated web applications, offering services such as 3D modeling, machine learning, and blockchain solutions. The company aimed to empower businesses with transformative technology to fuel growth, catering to both startups and enterprises. However, the official website, devverse.org, is currently inaccessible due to an expired domain. This means that details regarding its specific features, pricing, and current offerings are unavailable. Users interested in DevVerse's services would need to wait for the domain to be renewed to access any information about its AI-powered solutions.

TensorFlow-Object-Detection-on-the-Raspberry-Pi provides a comprehensive, step-by-step tutorial for implementing TensorFlow's Object Detection API on a Raspberry Pi. This guide enables users to perform real-time object detection on live video feeds from a Picamera or USB webcam. It includes updated instructions for easily installing TensorFlow and the protobuf compiler, making the setup process more accessible. The repository also features a 'Pet Detector' program as an example application, demonstrating how to use object detection to send text alerts when specific objects are detected. This tutorial is ideal for developers looking to create unique detection applications on the Raspberry Pi.

The Temporal Shift Module (TSM) is an open-source PyTorch implementation designed for efficient video understanding. It allows for temporal modeling in video analysis tasks, such as action recognition, by shifting part of the channels along the temporal dimension. TSM is a plug-and-play module that adds zero parameters and zero FLOPs, making it highly efficient. The project provides pre-trained models on datasets like Kinetics-400 and Something-Something, along with code for data preparation, testing, and training. It also features a live demo for online hand gesture recognition on NVIDIA Jetson Nano, showcasing its real-time capabilities.

YOLO-Patch-Based-Inference is a Python library designed to simplify SAHI-like inference for instance segmentation tasks, specifically enabling the detection of small objects in images. It caters to both object detection and instance segmentation, supporting various Ultralytics models including YOLOv8, YOLOv9, YOLOv10, YOLO11, YOLO12, FastSAM, and RTDETR. Users can leverage pre-trained models or integrate their custom-trained models. The library also provides extensive customization options for visualizing inference results, applicable to both standard and patch-based inference methods. It includes interactive notebooks and tutorials to guide users through batch inference procedures, custom visualization, and more.

3DMPPE_POSENET_RELEASE is the official PyTorch implementation of the 'Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image' presented at ICCV 2019. This repository specifically focuses on the PoseNet component of the system. It offers a flexible and simple codebase compatible with various 2D and 3D, single and multi-person pose estimation datasets, including Human3.6M, MPII, MS COCO 2017, MuCo-3DHP, and MuPoTS-3D. The tool also includes visualization code for human pose estimation, making it valuable for researchers and developers working on computer vision tasks related to human understanding. Users can train and test the network, and integrate their own datasets by converting them to MS COCO format.

AS-One is a comprehensive, open-source Python wrapper designed for computer vision tasks, providing an easy and modular interface for object detection, segmentation, tracking, and pose estimation. It supports a wide range of YOLO models, including YOLOv9, v8, v7, v6, v5, R, and X, enabling users to implement these advanced models in under 10 lines of code. The library integrates various tracking algorithms like ByteTrack, DeepSORT, and NorFair, and supports models in ONNX, PyTorch, and CoreML formats. AS-One also includes capabilities for text detection and recognition using models like CRAFT and EasyOCR, and pose estimation with YOLOv8 and YOLOv7-w6. It is ideal for developers and researchers looking for a unified and efficient solution for their computer vision projects.

FastDepth is an open-source project that provides trained models and evaluation code for fast monocular depth estimation, specifically optimized for embedded systems. The project includes resources for setting up the environment, downloading pre-trained models, evaluating performance, and deploying models on hardware like the NVIDIA Jetson TX2. It leverages PyTorch for model training and evaluation, and the TVM compiler stack for efficient cross-compilation and deployment. The repository offers various MobileNet-NNConv5 architectures, including pruned versions with additive skip connections, demonstrating significant performance improvements in terms of RMSE and delta1 metrics compared to prior work, with very low runtimes on embedded GPUs.

LuaRadio is a lightweight and embeddable flow graph signal processing framework specifically designed for software-defined radio (SDR). Built on LuaJIT, it offers a small binary footprint and no external hard dependencies, making it highly portable. The framework provides a comprehensive suite of source, sink, and processing blocks, along with a simple API for defining and running flow graphs, creating custom blocks, and managing data types. It's ideal for rapidly prototyping software radios, developing modulation/demodulation utilities, and conducting signal processing experiments. LuaRadio can also be embedded into existing radio applications, serving as a user-scriptable engine for advanced signal processing tasks. It supports computational acceleration through LuaJIT's FFI to wrap external libraries like VOLK, liquid-dsp, and others, ensuring efficient performance.

llm-twin-course is a free educational resource designed to guide users through the process of building a production-ready Large Language Model (LLM) and Retrieval Augmented Generation (RAG) system. The course emphasizes LLMOps best practices, offering practical, hands-on lessons and accompanying source code. It covers the entire development lifecycle, from initial data gathering to the final stages of productionizing LLMs, with a specific focus on creating an AI replica.

MY-BASIC is a lightweight BASIC interpreter implemented in standard C, provided in dual files for easy integration. It aims to be highly embeddable, extendable, and portable across various platforms. The interpreter supports dynamic typing, structured syntax, and a unique blend of prototype-based object-oriented programming with functional programming via lambda abstraction. Its core is compact, allowing it to be used as a standalone interpreter or seamlessly embedded into existing projects developed in C, C++, Java, Objective-C, Swift, C#, and more. Developers can customize its functionality by adding their own scripting interfaces, making it a versatile tool for various programming needs.

ObjectDetection-OneStageDet is an open-source object detection framework developed by Tencent, designed to provide a unified platform for single-stage generic object detectors. Currently, it supports YOLOv2 and YOLOv3 implementations, with future plans to integrate YOLO and SSD into a single framework. The tool emphasizes performance and speed, offering good mAP scores and fast inference times, especially with various efficient backbones like TinyYOLO, MobileNet, and ShuffleNet. It provides comprehensive instructions for installation, data preparation, training, evaluation, and benchmarking, making it suitable for developers and researchers working on object detection tasks.

Flow-Guided Feature Aggregation (FGFA) is an open-source implementation for video object detection, initially described in an ICCV 2017 paper. It offers an accurate and end-to-end learning framework, significantly improving object detection accuracy in videos, particularly for fast-moving objects, by aggregating nearby frame features along motion paths. The tool is end-to-end trainable for video object detection and includes motion-specific evaluation code to assess detection accuracy for slow, medium, and fast-moving objects. This repository is based on MXNet and was developed by interns at MSRA, building upon previous work like Deep Feature Flow.

The Edinburgh Centre for Robotics (ECR) is a leading research and training institution dedicated to advancing Robotics and Autonomous Systems (RAS). It brings together over 50 world-leading investigators from Heriot-Watt University and the University of Edinburgh. The ECR focuses on research topics related to safety and safe interaction between robots, people, and their environments, applying fundamental theoretical methods to real-world problems. The center also hosts Centres for Doctoral Training (CDTs) in AI, robotics, and autonomous systems, including the UKRI AI CDT in Dependable and Deployable AI for Robotics (CDT-D2AIR). These programs aim to produce innovation-ready graduates equipped with technical, scientific, ethical, and enterprise skills, aligning closely with industrial project partners across various RAS market sectors.

Rofunc is an open-source Python package designed for robot learning from demonstration and robot manipulation. It provides a comprehensive framework for developing and deploying advanced robot learning algorithms. The tool is hosted on GitHub, making it accessible for researchers and developers in the robotics field. Rofunc facilitates the entire workflow, from initial algorithm development to practical deployment, supporting various aspects of robot control and interaction. Its open-source nature encourages community contributions and collaborative development, making it a valuable resource for advancing robotics research and applications.

sqlite3-ruby offers Ruby bindings for the SQLite3 embedded database, enabling Ruby developers to seamlessly integrate and utilize the SQLite3 database engine within their applications. This library is designed for compatibility with SQLite 3.6.16 or newer, ensuring a broad range of support for existing and new projects. It facilitates database operations directly from Ruby code, making it a valuable tool for managing data in Ruby-based applications. The integration simplifies database interactions, allowing developers to focus on application logic rather than complex database setup.

Stereo-RCNN is an open-source implementation for accurate 3D object detection and estimation, primarily developed for autonomous driving applications. This tool leverages stereo images to perform simultaneous object detection and association, enhancing the precision of 3D box estimations. It also incorporates a dense alignment module for refining 3D box predictions. The project supports Pytorch 1.0.0 and Python 3.6, with a light-weight version available for scenarios with limited GPU memory. Researchers and developers can utilize Stereo-RCNN for tasks requiring robust 3D perception from image-only data, offering a valuable resource for advancing autonomous systems.

Turbo is a robust framework designed for LuaJIT 2, aimed at simplifying the development of fast and scalable network applications. It leverages an event-driven, non-blocking, and no-thread design to achieve excellent performance and a minimal footprint, making it suitable for high-load applications and embedded systems. The framework supports various network applications, including HTTP REST APIs, dynamic web pages via templating, and WebSockets. It provides generic building blocks like an I/O loop and IO Stream classes, along with customizable TCP (with SSL) server classes. Turbo is particularly optimized for the HTTP(S) protocol, catering to web and HTTP API developers, while also offering direct integration with existing C libraries for ultimate memory and CPU performance.

vedadet is a single-stage object detection toolbox built on PyTorch, offering a modular design that re-engineers MMDetection for enhanced flexibility and deployment. It decomposes the detector into four key parts: data pipeline, model, postprocessing, and criterion, making it straightforward to convert PyTorch models into TensorRT engines. This design facilitates efficient deployment on NVIDIA devices such as Tesla V100, Jetson Nano, and Jetson AGX Xavier. The toolbox supports several popular single-stage detectors, including RetinaNet and FCOS, right out of the box. Its friendly integration with TensorRT allows for easy model conversion and deployment through both Python and C++ front-ends, making it a powerful tool for developers working on object detection tasks.