Context The ability to monitor species in their natural habitat is useful when determining how the species respond to changes in environment. This may be particularly important when the species being studied are endangered and a population decline will trigger management action. The Eastern Ground Parrot (*Pezoporous wallicus wallicus*) is found only in Australia, where it is officially listed as a rare and threatened species. Furthermore, it is difficult to directly observe, as it lives in dense vegetation and rarely flies. As such, the chance of successful monitoring depends on locating the vocalisation or calls. One of the techniques that has been commonly used to monitor Ground Parrots (GPs) in their natural habitat is for experienced ecologists to listen to the calls and manually locate and plot the perceived distribution of the birds. More recently, surveys have also used techniques based on audio signal processing and pattern recognition, where records of acoustic signals are analysed and specific vocalisations are detected. Such detection can be used for the purpose of recognition and estimation of the number of GPs in the habitat. To facilitate research on bird call detection using pattern recognition techniques, we provide here a dataset of GP vocalisation. The dataset contains 4 audio sequences of different lengths sampled at 16kHz using Song Meter SM2 devices. The dataset was recorded at dusk and dawn times on different days and at four different locations in the Barren Grounds Nature Reserve, a protected nature park located in the Southern Highlands region of New South Wales, Australia. The sequences include overlapping GP calls and sounds from sources other than GP (e.g. wind, spurious noise, and vocalisations of other species). All the GP calls in the dataset are manually annotated and used as the ground truth for evaluation of detection algorithms. Content This is the dataset used for Ground Parrot Call detection project. The folder Data includes 1) 4 audio sequences formatted in wav files and named Sq1, Sq2, Sq3, and Sq4. 2) Training. This folder contains the training data and includes 2 sub-folders: GroundParrot (ground parrot calls) and Others (sounds other than ground parrot calls). Each folder contains 25 wav files and each file is 1 second length. 3) GroundTruth. This folder includes 4 files. Each file contains the ground-truth of the ground parrot calls of one of the 4 sequences, e.g. Sq1_gt.txt is the ground-truth of the sequence Sq1.wav. The ground-truth files are written in the following format, - The first line contains the number of ground parrot calls. - Each following line describes a call with the start and end time stamp. The time stamps are represented in hours:minutes:seconds Acknowledgements This work was supported by funding from the Office of Environment and Heritage, NSW Australia.