We rolled up our sleeves and devised a sophisticated solution using Optical Character Recognition (OCR) technology, complemented by an advanced post-processing routine. We implemented a custom object detection model sharp enough to identify player cards and associate them with the correct names, ensuring we captured all the necessary data for Pokerstats comprehensively and accurately. We customized our solution to handle videos from specific YouTube channels and set predefined formats for seamless data collection. To address the challenges of processing lengthy video sessions, we employed parallel processing. This modular approach in our application meant that one processing component, like OCR, didn’t have to wait for another, such as the object detector, to begin its task on the next frame. This allowed us to process multiple video frames simultaneously, speeding up data collection and ensuring no detail was missed.
