PySceneDetect is a production-grade Python library and CLI tool for detecting scene changes (cuts, fades, dissolves) in video files. It provides multiple detection algorithms (ContentDetector, AdaptiveDetector, ThresholdDetector, HistogramDetector, HashDetector, and TransNetV2 ONNX) and supports several video backends (OpenCV, PyAV, MoviePy). The architecture is modular: a VideoStream abstraction for decoding, a SceneDetector interface for pluggable algorithms, a SceneManager for orchestration, and output writers for CSV, EDL, FCP XML, OTIO, HTML, images, and video splitting. The CLI (built with click) offers chained subcommands for detection, splitting, and export. The repository also includes a comprehensive benchmark suite for evaluating accuracy on public shot-boundary datasets (BBC, AutoShot, ClipShots) using TRECVID-SBD metrics, making it valuable for both practical video processing and research. The codebase is well-structured, has extensive test coverage, and is actively maintained with recent releases and CI workflows.
MediaInfo is a mature, cross-platform application for extracting and displaying technical and tag metadata from audio and video files. It provides a command-line interface (CLI) and multiple graphical front-ends (Qt, wxWidgets, Cocoa, VCL, UWP, Android, iOS) all sharing a common core library (MediaInfoLib). The repository contains the full source for the CLI, all GUIs, Windows shell extension, installers, and build/packaging scripts. The architecture is a classic layered design: a central Core class wraps MediaInfoLib and is consumed by each GUI via platform-specific adapters. The codebase is functional and widely used, but suffers from significant fragmentation across GUI frameworks and build systems, with legacy components (VCL, UWP) and conditional compilation adding maintenance overhead. The translation system using CSV files is well-designed and the command-line parser is robust, but most components are tightly coupled to the MediaInfo domain and not easily reusable outside the project.
trdrop is a desktop application for analyzing uncompressed video files to estimate real frame rates, frame times, and detect screen tearing. Built with Qt/QML for the UI and C++/OpenCV for core processing, it supports simultaneous analysis of up to 3 videos. The application features a pipeline architecture: video capture → frame processing (difference calculation, framerate/frametime estimation, tear detection) → image conversion → composition → rendering (with configurable graphs and overlays) → export (image sequences or CSV). Key strengths include sophisticated tear detection logic that adjusts frame rate calculations for partial frame tears, clean separation between UI and processing logic, and dynamic resolution-adaptive rendering. The project is actively maintained with recent commits and a functional CI/CD pipeline for Windows builds.
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.