Coding & Development

Bineric AI offers secure, privacy-focused AI solutions designed for enterprise-grade security and compliance. The platform provides tools like Bineric Chat, which includes NorskGPT for sensitive use cases, and an AI Models and API Platform for developers to deploy AI anywhere. With a strong emphasis on privacy, Bineric ensures that AI deployments meet stringent security standards. It caters to organizations looking for robust AI infrastructure that prioritizes data protection and offers a range of AI models for various applications.

lite.ai.toolkit is a comprehensive C++ AI toolkit offering over 100 pre-trained AI models for a wide range of computer vision tasks. It supports multiple inference engines including MNN, ONNX Runtime (ORT), and TensorRT (TRT), allowing for flexible deployment across different platforms like Linux, macOS, Windows, and Android. The toolkit includes models for object detection (YOLO series, SSD, EfficientDet), face detection and recognition (RetinaFace, ArcFace), segmentation (MODNet, RobustVideoMatting), and other advanced functionalities like Stable Diffusion and Face Fusion. It emphasizes simplicity and user-friendliness with consistent syntax, minimal dependencies, and detailed build instructions for easy integration into C++ projects.

Latitude-llm is an open-source platform designed for building and operating LLM features in production. It emphasizes observability and evaluation, allowing users to instrument existing LLM calls to capture prompts, inputs/outputs, tool calls, latency, token usage, and cost. The platform supports a reliability loop that turns production failures into repeatable fixes through features like issue discovery, automatic evaluations, and a prompt optimizer. Users can start with observability and evaluations, then progress to a reliability loop to continuously improve prompts. Latitude-llm works with most model providers and frameworks out of the box and offers both a managed cloud product and a self-hosted deployment option.

Marian is an efficient open-source Neural Machine Translation framework implemented in pure C++ with minimal dependencies. It is designed for high performance, supporting fast multi-GPU training and GPU/CPU translation. The framework incorporates state-of-the-art NMT architectures, including deep RNN and transformer models, making it suitable for advanced machine translation research and development. Marian is released under a permissive MIT open-source license, encouraging broad adoption and contribution. Its focus on efficiency and C++ implementation provides a robust foundation for building and deploying neural machine translation systems.

Deep-active-learning is an open-source Python library designed for implementing and experimenting with various active learning algorithms. It provides a collection of methods such as Random Sampling, Least Confidence, Margin Sampling, Entropy Sampling, Uncertainty Sampling with Dropout Estimation, Bayesian Active Learning Disagreement, Cluster-Based Selection, and Adversarial Margin. This library is particularly useful for researchers and developers in the field of machine learning who aim to reduce the amount of labeled data required for training models while maintaining or improving performance. The repository includes prerequisites and a demo script for easy setup and experimentation, making it a practical tool for exploring active learning strategies.

Data-Science-Projects is an open-source GitHub repository offering a comprehensive collection of data science projects. Each project is meticulously organized within its own directory, containing all necessary code, relevant datasets, detailed documentation, and additional resources. The repository covers a wide array of topics, including various prediction models such as Breast Cancer Prediction, Red Wine Quality Prediction, Heart Stroke Prediction, House Price Prediction, and many more. It serves as an excellent resource for students and developers looking to explore practical applications of machine learning, data analysis, and visualization techniques, providing concrete examples and results for each project.

daclip-uir provides an official PyTorch implementation for controlling vision-language models, specifically designed for universal image restoration tasks. This tool can address various image degradations such as motion blur, haze, JPEG compression, low-light conditions, noise, raindrops, rain, shadows, snow, and uncompleted images (inpainting). It offers pretrained models for degradation-aware CLIP and universal image restoration, along with a Gradio app for easy testing of custom images. The project also includes a follow-up work focusing on photo-realistic image restoration and handling real-world mixed-degradation images, demonstrating its continuous development and robust capabilities in the field.

deep-image-retrieval is an open-source project from Naver Labs Europe focused on advancing image retrieval through deep learning. It offers models and evaluation scripts implemented in Python3 and PyTorch 1.0+, enabling researchers and developers to learn deep visual representations for image retrieval tasks. The tool supports training image retrieval systems using various loss functions, including triplet loss and a novel Average Precision (AP) loss, which directly optimizes for retrieval performance. It includes pre-trained models based on Resnet architectures with different pooling mechanisms (MAC, GeM) and provides scripts for evaluating these models on standard benchmarks like Oxford5K and Paris6K, as well as for extracting features from custom image datasets.

Deep-Learning-Approach-for-Surface-Defect-Detection is an open-source project offering a Tensorflow implementation of a segmentation-based deep learning approach for surface defect detection. This tool is designed for automated visual inspection and quality control, particularly relevant in manufacturing processes. It allows users to train a deep learning model on datasets like KolektorSDD to identify and classify surface imperfections. The implementation supports independent training of segmentation and decision networks, providing flexibility for model optimization. It includes scripts for testing, training, and visualization of results, making it a practical resource for researchers and developers working on computer vision applications for industrial quality assurance.

deep-learning-models is a GitHub repository offering Keras code and pre-trained weights for several widely used deep learning models. This resource includes implementations for VGG16, VGG19, ResNet50, Inception v3, and a CRNN for music tagging. The architectures are designed to be compatible with both TensorFlow and Theano backends, automatically adapting to the image dimension ordering specified in your Keras configuration. Users can easily load pre-trained weights, such as 'imagenet' for image models or 'msd' for the music tagging model, which are automatically downloaded and cached locally. While this repository is deprecated in favor of `keras.applications`, it remains a valuable reference for understanding and utilizing these foundational models.

Deep-Learning-TensorFlow is a GitHub repository offering a collection of pre-built Deep Learning algorithms implemented with the TensorFlow library. This package is designed as a command-line utility, enabling users to quickly train and evaluate popular Deep Learning models. It can also serve as a benchmark or baseline for comparing custom models and datasets. The repository includes implementations for Convolutional Networks, Restricted Boltzmann Machines, Deep Belief Networks, Deep Autoencoders, Denoising Autoencoders, Stacked Denoising Autoencoders, and MultiLayer Perceptrons. It also supports Logistic Regression. The package can be installed via pip as 'yadlt' or by cloning the GitHub repository, and it features a scikit-learn-like interface for ease of use.

node-tensorflow is an open-source project offering a NodeJS API for Google's powerful machine learning library, TensorFlow. This initiative focuses on making TensorFlow's capabilities readily accessible to JavaScript developers within the NodeJS environment, prioritizing both performance and stability. Currently in its early design stages, the project actively seeks contributions, particularly from individuals with C++ knowledge, to accelerate its development. It leverages SWIG for interfacing the C++ core API with Node.js bindings, with a roadmap that includes integrating Python API features like Optimizers and Tensor Transformations. The goal is to evolve into a robust Node.js API for end-users, enabling them to build and control TensorFlow computation graphs directly from Node.js.

DeepReg is a freely available, community-supported open-source toolkit designed for research and education in medical image registration using deep learning. It is built on TensorFlow 2 for efficient training and rapid deployment of models. The toolkit implements major unsupervised and weakly-supervised algorithms, along with their combinations and variants, focusing on growing and diverse clinical applications. All DeepReg Demos utilize openly accessible data, and it offers simple built-in command-line tools that require minimal programming. DeepReg operates under the Apache 2.0 license, promoting an open, permissible, and research-and-education-driven environment.

DeepSeek R1 Online is a revolutionary open-source AI model designed for advanced reasoning, mathematics, and coding tasks. It offers free, no-login access and boasts capabilities comparable to leading proprietary solutions like OpenAI o1, often at a significantly lower cost. Built on a sophisticated Mixture of Experts (MoE) architecture with 37B active/671B total parameters and 128K context length, DeepSeek R1 implements advanced reinforcement learning for self-verification and multi-step reflection. It achieves state-of-the-art performance on benchmarks like MATH-500 (97.3% accuracy) and Codeforces (96.3% percentile). The tool also supports local deployment via WebGPU and offers various distilled models for resource-constrained environments.

ModelingToolkit.jl is a high-performance symbolic-numeric computation framework designed for scientific computing and scientific machine learning within the Julia ecosystem. It allows users to define models at a high level, enabling symbolic preprocessing for analysis and enhancement. The tool can automatically generate optimized functions for model components, such as Jacobians and Hessians, and automatically sparsify and parallelize computations. It also applies automatic transformations, like index reduction, to simplify models for numerical solvers. ModelingToolkit.jl supports composing multiple ODE subsystems and simulating complex Differential-Algebraic Equations (DAEs), making it a powerful tool for advanced scientific modeling and simulation.

DeepLearningZeroToAll is an open-source project offering a comprehensive collection of TensorFlow basic tutorial labs. It provides practical code examples designed to help users understand fundamental deep learning concepts. While the current tutorials are primarily in Korean, there are plans to release English video tutorials, making it accessible to a broader audience. The project emphasizes readability and understandability over efficiency, making it an excellent resource for instructional purposes. It covers various deep learning topics, including linear regression, logistic regression, softmax classifiers, CNNs, and RNNs, with examples implemented in TensorFlow, Keras, MXNet, and PyTorch. The repository encourages community contributions and provides guidelines for code style and testing.

DnCNN is a deep convolutional neural network designed for various image restoration tasks, primarily focusing on image denoising. It leverages residual learning to effectively remove additive white Gaussian noise (AWGN) from images. The tool is implemented in PyTorch and MatConvNet, offering flexible training and testing options. Beyond denoising, DnCNN can also be applied to single image super-resolution (SISR) and JPEG image deblocking, demonstrating its versatility. The architecture benefits from batch normalization and residual learning, which stabilize training and allow a single model to handle different tasks. It provides state-of-the-art performance in Gaussian denoising and is available as open-source code on GitHub.

Dobb·E is an open-source, general framework designed for learning household robotic manipulation. It provides both hardware and software components, including 3D printable STL files for a demonstration collection tool called "The Stick," and software for processing collected data. The framework also includes code for training policies using pretrained models and deploying learned policies on robots. Dobb·E aims to enable robots to learn new tasks with minimal user demonstration, leveraging a dataset of 1.5 million RGB-D frames collected in various home environments. This initiative seeks to accelerate research in home robotics and address unique challenges encountered in real-world household settings.

DI-engine is a generalized decision intelligence engine built for PyTorch and JAX, offering a comprehensive framework for reinforcement learning. It features python-first and asynchronous-native task and middleware abstractions, integrating key decision-making concepts like Env, Policy, and Model. The framework supports a wide array of deep reinforcement learning algorithms, including DQN, PPO, SAC, and multi-agent, imitation, offline, and model-based RL. Beyond algorithms, DI-engine aims to standardize decision intelligence environments and applications, catering to academic research and prototype development. It also includes highly re-usable modules for RL optimization, PyTorch utilities, and system optimizations for efficient large-scale RL training.

DR Tulu is an open-source Deep Research (DR) model designed for tackling long-form research tasks. The DR Tulu-8B model has demonstrated performance comparable to OpenAI DR on long-form DR benchmarks. This repository provides the official code for DR Tulu, including an agent library with a MCP-based tool backend, high-concurrency async request management, and a flexible prompting interface for developing and training deep research agents. It also includes RL training code based on Open-Instruct and SFT training code based on LLaMA-Factory, allowing for supervised fine-tuning and reinforcement learning with GRPO and evolving rubrics. An interactive CLI demo is available for users to experiment with DR Tulu-8B.

Float16 is a comprehensive GPU management platform designed for deploying, managing, and scaling AI models. It offers a full spectrum of services including AI-as-a-Service (AaaS) for instant access to ready-to-use AI models without coding, Platform-as-a-Service (PaaS) for flexible resource allocation, and Infrastructure-as-a-Service (IaaS) for bare-metal GPU instances. The platform emphasizes ease of use with one-click deployment, significantly reducing setup time from weeks to minutes. Float16 provides dedicated and isolated GPU resources, ensuring zero interference and optimal performance for workloads. It features a credit-based quota system for flexible GPU utilization, eliminating waste from fixed time slots. Supported by NVIDIA Inception Program, Float16 is ideal for ML engineers, data scientists, software developers, and researchers seeking efficient and scalable GPU solutions.

dissecting-reinforcement-learning is an open-source repository offering Python code, PDFs, and supplementary resources for a series of blog posts on Reinforcement Learning. It serves as a comprehensive guide for practitioners and students, covering fundamental concepts like Markov chains, Bellman Equation, Monte Carlo methods, and Temporal Difference Learning. The repository also delves into advanced topics such as Actor-Critic methods, Evolutionary Algorithms, and various function approximation techniques including neural networks. It provides standalone Python environments for classic RL problems like the Inverted Pendulum, Mountain Car, and Multi-Armed Bandit, which do not require external installations like OpenAI Gym. This makes it an accessible resource for hands-on learning and experimentation.

DLFS_code is a GitHub repository containing all the code from the book "Deep Learning From Scratch," published by O'Reilly in September 2019. It is designed for readers to clone and systematically step through the code to better understand the deep learning concepts presented in the book. The repository is structured by chapter, with each chapter featuring two notebooks: a Code notebook with runnable Python code and a Math notebook for LaTeX equations. It includes implementations of deep learning models, such as a single-layer CNN trained from scratch in pure Numpy to achieve over 90% accuracy on MNIST, as detailed in the book's Appendix.

FunctionGemma Tuning Lab offers a Gradio-based web interface designed for fine-tuning FunctionGemma models, specifically for tool calling applications. This open-source tool, available on Hugging Face, allows developers and data scientists to customize the FunctionGemma model to suit their specific needs. The interface supports multiple users simultaneously, processing requests in a queue to ensure efficient interaction. Licensed under Apache 2.0, it provides a flexible and accessible platform for experimenting with and adapting FunctionGemma for various projects and use cases, making advanced model tuning more approachable.