Show HN: Real time visual saliency detection (github.com)

Dosage is a new Python library designed for real-time visual saliency detection on images and videos, offering a practical tool for researchers and developers interested in understanding which parts of an image or video frame capture the most visual attention. Currently compatible with macOS and Linux, it can potentially run on Windows via the Windows Subsystem for Linux (WSL). The library is easy to install via pip and includes example scripts for both real-time processing and saving processed video outputs to disk. This open-source project is significant for the AI/ML community because real-time saliency detection has applications in fields like computer vision, human-computer interaction, and content optimization. By enabling on-the-fly detection of salient visual regions, Dosage can support more efficient model training, video compression strategies, or interactive systems that adapt to user focus. The availability of ready-to-use examples lowers the barrier for experimentation, allowing developers to quickly integrate and test saliency detection in their workflows. Technically, Dosage processes video streams with minimal latency, providing immediate feedback on which features stand out visually. This real-time capability makes it valuable for dynamic scenarios such as live video analysis or adaptive rendering, where understanding viewer attention instantaneously can drive smarter AI-powered decisions. Overall, Dosage contributes a streamlined, accessible approach to incorporating saliency-aware insights into AI and computer vision projects.