(objects and detection targets) Types of data used Infrastructure Railway inspection Inspection of train railway Image data including the target construction Infrastructure Scrap copper estimation Estimation of scrap copper (breaking down to subdivisions) Image data per type Infrastructure Estimation of danger levels Objects/positions where incidents may happen (building exits, etc.) Image data, sensor data Logistics Estimation of anomaly Detect anomaly of logistics equipment Sensor data and anomaly data of the target object to be detected Logistics Visualize packaging Measurement of the time spent for packaging and the number of packed items Video of packing work of the target products, image of the target items to be packaged, classification linked to the items to be packaged Logistics Prevent dangerous driving Raise an alert for dangerous driving based on multiple definitions Data of drivers obtained by sensors Construction, Logistics Risk/incident mitigation Raise an alert when people come too close to machines Video data of the target in the factory to be detected Game Cheat detection Detect unauthorised accounts from action log in a game. Auto monitoring of field activity log, etc., by AI to check/monitor accounts conducting unauthorised actions. Action log in a game, battle history, field activity log SNS Detect unauthorized posts (social media function) Auto monitoring by AI to check if posted texts, images, videos, etc., by users include inappropriate contents (abusive language, offensive to public order and morals) Image, video, texts Media Advertisement CTR prediction Use past Advertisement distribution data to train image features and actual results to predict the degree of effectiveness by the features of new images Image Advertisement data (image file, number of display, number of clicks, number of conversion) Media 1to1 recommendation Optimize display of contents in the client’s service User data (purchase log, user master data, feedback data), Metadata Communicatio n Demand prediction for store sales Demand prediction for the number of store sales per store/ per product Past purchase data per store, customer data per store, Factor data that may correlate with demand per store Leasing Total platform Maximize profitability by dynamic pricing Past sales history data, stock status data, external input data (Big data - events, weather, competitors, social media, etc.) Medical Automated medical inspection Automated ocular fundus examination by utilizing AI Medical image data Confidential 16