tensorflow_object_counting_api

Why we included this project

Engineers building object counting systems from video feeds will find a practical reference here. The API packages detection, tracking, and counting into straightforward calls, so you can process a traffic or pedestrian clip and get cumulative totals across a region-of-interest line, or follow a single class like vehicles in real time. Ready-to-run examples for vehicle and pedestrian counting make it easy to see a full pipeline before adapting it to your own data. A custom mode demonstrates transfer learning for training counters on your own classes, and a Mask R-CNN variant adds segmentation when you need it. The code predates recent TensorFlow releases, so expect to modernize the APIs yourself, but as a self-contained example it remains one of the more accessible ways to build a vision-based counter.