Developing image-based automatic glaucoma diagnosis systems requires a large amount of fundus image groundtruth data, especially that from glaucomatous patients. We develop ACHIKO-K database which contains 258 manually annotated retinal images and their clinical diagnosis information. All images in ACHIKO-K are taken from glaucoma patients and contain rich information on glaucoma related pathological signs, e.g. hemorrhage, optic nerve drusen and optic cup notching etc. Using our in-house developed image marking tool, optic cup/disc boundaries and various pathological signs are carefully annotated, providing accurate groundtruth for the development of image based glaucoma detection algorithms. The metadata embedded in the digital image files are extracted and make available in ACHIKO-K database, which, possesses rich information on camera setting and are critical for developing preprocess components for image normalization. Furthermore, the fundus images in ACHIKO-K contains timely information recording retinal changes and can be precious resources for retrospective study.