olafenwamoses/imageai

ImageAI is a Python library that makes it possible to add computer vision features to an application with only a small amount of code. Computer vision refers to software that can analyze images or video to understand what is in them, such as identifying objects, classifying scenes, or tracking movement. The library covers four main capabilities. Image classification takes a single image and returns a label describing what it contains, using one of four pre-trained models that were trained on a dataset of 1,000 categories. Object detection goes further by finding and locating multiple distinct objects within an image, drawing bounding boxes around each one. Video object detection applies the same idea to video, processing frames in sequence to identify objects as they appear and move. Finally, ImageAI lets you train your own custom classification model using your own labeled images, so you can teach it to recognize categories that are not in the default dataset. The design goal is to reduce the code needed to do these tasks. Each feature is accessible through a few lines of Python: load a model, point it at an image or video file, and get results back. You do not need to understand the underlying machine learning architecture to use the library. Installation is done through pip, the standard Python package installer. The library can run on a CPU for lower-speed use, or on a GPU with CUDA support for faster processing. Compatible Python versions are 3.7 through 3.10. The README also mentions two newer projects from the same author (Jarvis and TheiaEngine) that offer expanded capabilities, but those are separate products. The full README is longer than what was shown.