FaithC is a Python library (with CUDA acceleration) for near-lossless 3D mesh encoding and decoding using Faithful Contour Tokens (FCT), as accepted at CVPR 2026 (Oral). It implements an encoder that transforms a mesh into compact per-voxel tokens (anchors, normals, edge flux signs) via octree traversal, SAT-based triangle clipping, QEF optimization, and edge flux computation. The decoder reconstructs the mesh from tokens using edge-based quad extraction and triangulation. The library is not a standalone application but a research-oriented tool intended for downstream tasks like VAE and diffusion models. It provides both a modern class-based API (FCTEncoder/FCTDecoder) and an older functional API. The codebase is well-structured with modular components, custom CUDA kernels for performance, and a PyTorch-centric design. However, it has some maintenance risks including dual code paths, hardcoded CUDA dependencies, and a custom 'atom3d' package without version pinning.
ASRT (Automatic Speech Recognition Toolkit) is a Chinese speech recognition system that provides both training and inference capabilities. It uses a deep CNN acoustic model with Connectionist Temporal Classification (CTC) and an N-gram language model based on a maximum entropy hidden Markov model. The system supports multiple backends (TensorFlow/Keras and experimental PyTorch) and can be trained on several Chinese speech corpora (THCHS30, ST-CMDS, etc.). It includes HTTP and gRPC servers for real-time inference, a speech recorder utility, and client libraries. The architecture is monolithic with separate scripts for training, evaluation, and serving. The codebase is functional but shows signs of technical drift, such as using deprecated Keras APIs (fit_generator) and hardcoded paths. The language model implementation is a notable highlight, offering a non-trivial statistical decoding approach. The repository is popular but has a single maintainer and a high number of open issues.
GPT-SoVITS is a production-grade text-to-speech and voice conversion system with a Gradio-based WebUI and FastAPI REST API. It supports zero-shot voice cloning from a 5-second reference audio and few-shot fine-tuning with 1 minute of data. The system integrates multiple model versions (v1–v4, v2Pro, v2ProPlus) with autoregressive (GPT-style) and non-autoregressive (SoVITS/CFM) components, plus a BigVGAN vocoder. It includes a full pipeline for dataset preparation (voice separation, ASR, text labeling), training (with DDP, LoRA, DPO), and inference (streaming, parallel, ONNX/TorchScript export). The codebase is heavily modular with reusable text frontends for Chinese, Japanese, Korean, Cantonese, and English, as well as audio processing tools (UVR5, BS-Roformer, AP-BWE). Despite its breadth, the code exhibits significant technical debt: version sprawl, hardcoded paths, global state, and experimental hacks. It is actively maintained with frequent bug fixes and feature additions.
This is a Python application and library for singing voice conversion and real-time voice changing, forked from so-vits-svc and rebuilt around PyTorch, PyTorch Lightning, and a Click-based CLI. The actual code provides two user-facing entry points: the `svc` CLI with commands for inference, realtime microphone conversion, dataset preprocessing, training, clustering, and cleanup, and the `svcg` PySimpleGUI desktop application for interactive and realtime inference. Inference loads a HuBERT content encoder from Hugging Face, extracts pitch using CREPE, DIO, Harvest, or Parselmouth, then runs a VITS-style SynthesizerTrn with either HiFi-GAN or MS/MB-iSTFT decoders conditioned on speaker embeddings. Despite fastapi being listed as a dependency, no HTTP API server exists in the source; the project is a desktop/CLI tool, not a web service.
BeatNet is a Python library for real-time and offline music beat, downbeat, tempo, and meter tracking. It provides a complete system comprising a CRNN-based neural network for feature extraction, particle filtering for causal inference, and DBN for non-causal inference. The library supports four operational modes: streaming (microphone input), real-time (file processing), online (fast file processing), and offline (DBN-based). Beyond inference, it includes a complete training pipeline for training the CRNN model on custom datasets with data preparation tools, dataset handling, and validation procedures. The architecture is modular with clear separation between feature extraction, neural network modeling, and inference algorithms.
P-HAR is a video analysis application that performs human action recognition specifically tailored for adult content. It processes input videos through three complementary modalities: visual frames (using TimeSformer), human pose sequences (using PoseC3D), and audio features (using a ResNet-based model). The system extracts features from each modality, runs inference through specialized deep learning models, and fuses the results using a weighted late fusion approach that handles heterogeneous output spaces. The architecture follows a modular pipeline: video input undergoes frame extraction and preprocessing, then each modality is processed independently through its respective model pipeline. The RGB stream processes spatiotemporal visual features, the skeleton stream first performs human detection and pose estimation before action recognition, and the audio stream extracts loudness-normalized spectral features. Results from all three streams are aligned to a common label space and combined using weighted averaging to produce final action predictions. Output can be generated as an annotated video with overlayed action labels and confidence scores, or as a JSON file containing temporal predictions for further processing. Key entry points include the multimodal demo script for end-to-end video processing, utility scripts for dataset preparation and model evaluation, and specialized tools for audio filtering and data pipeline management. The application relies on the MMAction2 framework ecosystem (including MMDetection and MMPose) for core computer vision capabilities, supplemented by standard Python libraries for media processing and data handling.
This repository is the companion code for the O'Reilly book 'Hands-On Large Language Models' by Jay Alammar and Maarten Grootendorst. It provides 12 chapters of Jupyter notebooks covering fundamental and advanced LLM topics including tokenization, embeddings, text classification, clustering, prompt engineering, semantic search, multimodal models, fine-tuning, and generation. All notebooks are designed to run on Google Colab with free GPU resources and rely heavily on Hugging Face libraries (transformers, datasets, sentence-transformers, TRL) and external APIs (OpenAI, Cohere). The code is educational in nature, with clear step-by-step explanations and visualizations. While it is not a production application or a novel research pipeline, it serves as a high-quality, reproducible tutorial resource for practitioners learning LLM workflows. The repository includes pinned dependencies, Conda environment setup, and local installation guides, making it relatively easy to reproduce in a controlled environment, though API keys are required for some sections.
This repository is a comprehensive PyTorch library for person re-identification (re-ID), providing a full pipeline for training, evaluating, and deploying deep learning models. It includes a modular architecture with data managers (ImageDataManager, VideoDataManager), a large model zoo (OSNet, ResNet, DenseNet, HACNN, etc.), training engines (softmax, triplet, with video support), loss functions (cross-entropy with label smoothing, triplet with hard mining), metrics (CMC, mAP with Cython acceleration), and utilities (feature extraction, model export to ONNX/OpenVINO/TFLite, re-ranking). The codebase is well-organized into subpackages (torchreid.models, torchreid.data, torchreid.engine, etc.) and includes research projects (attribute recognition, deep mutual learning, neural architecture search). While the library is feature-rich and demonstrates strong engineering for its domain, it has not been actively maintained since mid-2022, relies on outdated dependencies (Python 3.6, torch 1.10), and contains deprecated API usage. It remains a valuable reference for re-ID practitioners but requires updates for modern environments.
The code answers the research question of how to customize large text-to-image diffusion models with minimal data and minimal parameter updates, and it compares against DreamBooth and Textual Inversion. It includes reusable components for few-shot personalization, delta checkpoint handling, lightweight model compression, multi-concept weight composition, and CLIP/DINO evaluation, although the codebase shows research-level roughness: duplicated CompVis/diffusers paths, hardcoded paths, and several known bugs in conversion and composition utilities.
bertviz is a Python library for interactive visualization of attention in Transformer-based NLP models. It provides three complementary views: Head View (per-head attention patterns), Model View (attention flow between tokens across layers), and Neuron View (query/key interactions within individual attention heads). The library is designed for Jupyter notebooks, rendering D3.js-based interactive HTML directly in cell output. It accepts attention tensors from most HuggingFace models via output_attentions=True, and includes a vendored, modified copy of older HuggingFace transformers code to expose query/key vectors for the Neuron View on BERT, GPT-2, RoBERTa, XLNet, and similar architectures.
Microsoft Magma is a research codebase for a single large multimodal vision-language-action (VLA) foundation model, as opposed to a general-purpose application. The core is an 8B-parameter model built on LLaMA-3-8B-Instruct with a ConvNeXt vision encoder, capable of understanding images/video and producing both textual responses and discretized robot/UI actions. The repository contains the full training stack: model definitions, data preprocessing, dataset-specific constructors, mixed RLDS/PyTorch data loaders, action tokenization, DeepSpeed training configs, and evaluation hooks for robotics benchmarks like LIBERO and SimplerEnv. It also ships FastAPI servers and demo agents for UI automation and robot trajectory planning, but these are secondary to the model training and data pipeline core, making this a hybrid repo with a dominant data-research character. The data lineage is central to the project. Raw heterogeneous datasets—Open X-Embodiment, Ego4D, EpicKitchen, SeeClick, LLaVA, and the in-house Magma-820K—are normalized into a unified multimodal conversation format via per-dataset Constructor subclasses. Visual point tracks from CoTracker2 are clustered, quantized, and rendered as 'Trace-of-Mark' overlays; bounding boxes and labels are rendered as 'Set-of-Mark' overlays. Continuous robot actions are discretized into token IDs aligned with the LLM vocabulary, allowing a standard causal LM objective to train action prediction alongside visual understanding. The resulting model can be deployed via the included FastAPI server for interactive inference. Reproducing the full pipeline requires access to gated or proprietary datasets, CoTracker2 checkpoints, a custom transformers fork, and significant GPU resources, so end-to-end reproduction is not turnkey, though inference on the released HuggingFace checkpoint is more accessible.
This repository implements Dinomaly, a reconstruction-based unsupervised anomaly detection method for industrial inspection datasets. It uses vision transformers (primarily DINOv2) as feature extractors with a noisy bottleneck and linear attention mechanism. The code supports both class-separated and unified training across MVTec-AD, VisA, and Real-IAD datasets. Key components include dataset loaders, custom optimizers, training utilities, and evaluation metrics. While the research code achieves state-of-the-art results, it contains hardcoded paths, version-specific dependencies, and dataset-specific logic that limit immediate reproducibility.
Craftium is a voxel-based game engine (Luanti/Minetest fork) extended with reinforcement learning capabilities. The core system provides a moddable 3D world with block-based gameplay, Lua scripting API, client-server architecture, and procedural map generation. The craftium add-on supplies Python bindings, Gymnasium environment interfaces, and RL training scripts to enable machine learning research within the game world. The system supports both traditional gameplay and AI experimentation through environment-specific mods that define reward functions, observations, and termination conditions for agents.
EchoTorch is a PyTorch-based research library for reservoir computing and echo state networks (ESNs). It implements multiple ESN variants, including a self-predicting ESN (SPESNCell), conceptor-based pattern manipulation with logic operations (AND, OR, NOT, PHI), a custom TimeTensor time-series wrapper, and supporting datasets such as Lorenz, Roessler, Mackey-Glass, NARMA, and MNIST. The repository is a library rather than an application: there are no web servers, CLIs, or service entry points; the primary interface is the Python package `echotorch` with submodules for neural network cells, datasets, transforms, costs, and utilities.
This repository implements Geometry Forcing, a research framework for video generation with geometric consistency. The core contribution marries video diffusion models (DFoT - Diffusion Forcing Transformer) with 3D representation learning (VGGT) to produce temporally consistent videos conditioned on camera poses. It includes full training pipelines (Hydra + PyTorch Lightning), a suite of video metrics (FVD, FID, LPIPS, VBench), and evaluation scripts for reprojection/revisit error using DROID-SLAM. The codebase is structured as a research template with reusable base classes for algorithms, data modules, experiments, and metric registries. Raw datasets (RealEstate10K, Kinetics-600, Minecraft) are downloaded and preprocessed into video clips with associated camera parameters. Transformations include VAE encoding for latent diffusion, augmentation (flip, reverse), and sliding-window clip extraction. The primary research question is whether video diffusion can be constrained by explicit 3D geometry to reduce inconsistencies in generated videos, measured by geometry-aware metrics. The repo also demonstrates composable sampling via History Guidance and an alignment loss that projects diffusion latents into VGGT feature space. While the code is well-engineered for research (modular design, config-driven, metric registry), it carries typical research code baggage: hardcoded paths, vendored third-party code (DROID-SLAM, VGGT), potential dependency pinning conflicts, and some dead/commented code. Reproducibility is moderate: dependencies are pinned but no environment definitions beyond requirements.txt, and dataset downloads rely on fragile YouTube scraping.
This repository is a curated collection of baseline solutions for data science competitions, primarily from Chinese platforms (科大讯飞 AI开发者大赛, Tianchi, DataFountain, AIWIN, etc.) and some international ones (Kaggle). It contains hundreds of Jupyter notebooks and Python scripts covering a wide range of tasks: structured data classification/regression (LightGBM, XGBoost), computer vision (ResNet, EfficientNet, YOLO, CRNN), natural language processing (BERT, TF-IDF, Qwen), time series forecasting, and multimodal challenges. Each baseline is self-contained, providing a quick-start pipeline from data loading to submission generation. The repository serves as a practical learning resource for competition participants, demonstrating common feature engineering techniques, model training loops, and ensemble strategies. However, the code quality is uneven: many notebooks are exploratory with hardcoded paths, outdated library versions, and minimal error handling. The repo lacks centralized dependency management, environment definitions, or modular reusable libraries, making reproducibility low without manual adaptation. Despite these limitations, it offers a valuable catalog of patterns and templates that can be adapted for new competitions.
This is an implementation of SV2TTS (Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis) with a real-time vocoder, structured as a three-stage deep learning pipeline: speaker encoder, Tacotron-based synthesizer, and WaveRNN vocoder. The system clones a voice from a few seconds of reference audio and then synthesizes arbitrary text in that voice. The architecture is a monolithic Python application with CLI and PyQt5 GUI entry points. The encoder module implements a GE2E-loss LSTM speaker encoder trained on VoxCeleb/LibriSpeech. The synthesizer module is a Tacotron variant conditioned on speaker embeddings. The vocoder module is a modified WaveRNN (fatchord version) that generates audio in real-time. All three stages have independent training scripts and inference interfaces. The codebase was a landmark open-source voice cloning project but is now significantly outdated. It pins Python 3.9, PyTorch 1.10, and uses deprecated APIs (scipy.ndimage.morphology.binary_dilation, visdom, webrtcvad). The synthesizer uses a pure-Python instead of CUDA-optimized attention mechanism. The deepmind_version.py vocoder references non-existent utils modules, suggesting incomplete refactoring. The CI pipeline only runs import smoke tests.
This repository is a large model zoo (500+ models) focused on converting and quantizing pre-trained computer vision models for edge deployment across multiple frameworks (TFLite, ONNX, OpenVINO, CoreML, TensorRT). It provides hundreds of download scripts for pre-converted model files, conversion pipelines (PyTorch/ TensorFlow to ONNX/TFLite), and a modest set of inference demo scripts for real-time video processing. The repo is primarily a research/engineering resource for edge AI practitioners, not a software application. It exhibits heavy duplication across model directories, relies on an external S3 bucket for model distribution, and uses older framework versions (TensorFlow 1.x/2.x, ONNX opset 11-13). Despite these maintenance concerns, the systematic conversion workflow and post-processing ONNX subgraph generators (e.g., NMS, box decoding) represent reusable patterns for anyone building end-to-end inference pipelines. The repo is hybrid in nature: while classified as data_research due to its focus on pre-trained model artifacts and lack of original training code, it also contains application-like demo code and CLI entry points.
Renate is a Python library for continual learning that provides implementations of various algorithms (Experience Replay, DER, Super-ER, etc.) to mitigate catastrophic forgetting when updating neural networks on sequential data. It features a modular architecture separating models, data handling, update strategies, memory buffers, and drift detection, enabling users to customize continual learning pipelines. The framework supports local execution and cloud deployment via Amazon SageMaker, includes hyperparameter optimization through Syne Tune, and offers benchmarking tools for evaluating continual learning methods across diverse datasets and scenarios. Key reusable components include a memory buffer system with reservoir and class-balancing strategies, Maximum Mean Discrepancy (MMD) based drift detectors for monitoring data distribution shifts, and a composable loss component system for constructing complex continual learning objectives. The library integrates with PyTorch Lightning for training and provides abstract base classes for models and data modules to facilitate extension. Although the project is explicitly archived and no longer maintained, the codebase demonstrates solid engineering practices with well-defined interfaces and decoupled components. It has been tested on vision and NLP benchmarks (CIFAR, CLEAR, DomainNet, Hugging Face datasets) and includes synthetic scenario generators for controlled experimentation. The archival status indicates that users should expect no further updates or security patches, but the core functionality remains usable for continual learning research and applications.
Flowframes is a Windows desktop application for AI-based video frame interpolation. It provides a GUI (WinForms) to orchestrate interpolation using models like RIFE, DAIN, and FLAVR, running on CUDA (PyTorch) or NCNN (Vulkan) backends. The core pipeline extracts frames from input videos via FFmpeg, runs AI inference to generate intermediate frames, and re-encodes the result. It supports batch processing, resume from interruption, various output formats (video, GIF, image sequences), and hardware acceleration detection. The architecture is a monolithic C# application that spawns external processes (FFmpeg, Python scripts, NCNN binaries) and manages their lifecycle. The codebase includes a legacy CodeLegacy folder with near-duplicate files, indicating incomplete refactoring. It also bundles a large vendored Python environment (pip, setuptools, etc.) for the AI runtimes, which introduces maintenance overhead. Despite these issues, the application is functional and actively maintained, with useful utilities like a Fraction struct for rational FPS arithmetic and a rolling average class for performance monitoring.