ahmetozlu/tensorflow_object_counting_api

Built on TensorFlow's object detection models, the API provides multiple counting modes—cumulative (crossing region-of-interest thresholds), real-time (per-frame tallies), and tracking-based approaches—with optional color recognition for object classification. It supports both standard detection pipelines and advanced segmentation via Mask R-CNN, enabling instance-level counting. Transfer learning capabilities allow training custom detectors on domain-specific data, as demonstrated in included sample projects.

1,332 stars. No commits in the last 6 months.