OlafenwaMoses/ImageAI

Supports multiple state-of-the-art architectures including MobileNetV2, ResNet50, InceptionV3, and DenseNet121 for image classification, plus RetinaNet and YOLOv3 for object detection trained on COCO datasets. Built on PyTorch backend, it enables real-time video object detection and tracking alongside custom model training for domain-specific detection tasks. The library abstracts complex deep learning workflows into minimal-code APIs while offering both CPU and GPU/CUDA inference options.

8,864 stars and 4,642 monthly downloads. No commits in the last 6 months. Available on PyPI.