In the field of image anomaly detection tasks, benchmark data such as MVTec AD is widely used. These are datasets from the manufacturing industry, and the imaging environment is uniform. On the other hand, in anomaly detection for outdoor infrastructures, the imaging environment cannot be uniform. In this study, we propose a dataset of outdoor concrete blocks (Gogan dataset) which includes a variety of imaging environments. Then, we applied existing anomaly detection methods to the MVTec AD and Gogan datasets under the same conditions and compared their performance. Based on the results, we discuss the issues involved in anomaly detection under conditions where the imaging environment cannot be unified.