The paper introduces the PanAf20K dataset, the largest and most diverse open-access video dataset of great apes in their natural environment. The dataset comprises over 7 million frames from approximately 20,000 camera trap videos collected at 14 field sites in tropical Africa as part of the Pan African Programme: The Cultured Chimpanzee. The footage is accompanied by rich annotations and benchmarks, making it suitable for various computer vision tasks such as ape detection and behavior recognition. The dataset aims to support AI analysis of camera trap information, which is crucial for conservation efforts given the endangered status of great ape species. The paper discusses the motivation for the dataset, its contribution, and provides an overview of the dataset's structure, including data acquisition and annotation methods. It also presents benchmark results for ape detection and behavior recognition using state-of-the-art models, highlighting the current limitations and future directions for improving performance. The dataset and code are available from the project website.The paper introduces the PanAf20K dataset, the largest and most diverse open-access video dataset of great apes in their natural environment. The dataset comprises over 7 million frames from approximately 20,000 camera trap videos collected at 14 field sites in tropical Africa as part of the Pan African Programme: The Cultured Chimpanzee. The footage is accompanied by rich annotations and benchmarks, making it suitable for various computer vision tasks such as ape detection and behavior recognition. The dataset aims to support AI analysis of camera trap information, which is crucial for conservation efforts given the endangered status of great ape species. The paper discusses the motivation for the dataset, its contribution, and provides an overview of the dataset's structure, including data acquisition and annotation methods. It also presents benchmark results for ape detection and behavior recognition using state-of-the-art models, highlighting the current limitations and future directions for improving performance. The dataset and code are available from the project website.