One of the most pressing worldwide issues is food waste. Food waste has increased dramatically as a result of population growth, fast urbanization combined with industrial development, and changes in lifestyles and economic positions. The study aims to develop a food inventory management and recipe recommendation system for perishable items stored in the refrigerator to lessen the food waste generated within a household. The system consists of a camera module, load cell, Raspberry Pi microcomputer, and a mobile application. The automatic addition of perishable items uses Convolutional Neural Network (CNN) for classifying fruits and vegetables through transfer learning method. Milk and meat products are manually added to the inventory. The mobile application is used for viewing the inventory, expiration of perishable items, and recipe recommendation. The recipe recommendation is based on the expiration and availability of perishable items to reduce household food waste. According to the gathered data, the precision of different classes of the fruit and vegetable model ranges from around 71 % up to 80 % in terms of its recall values, while the accuracy of the model is evaluated at 74 %. The recipe recommendation system was able to fully utilize the recorded perishables roughly 81 % of the time. For food waste reduction, a daily average of 108.4 g of perishable food waste, including food peels, leftovers, and expired ingredients, has been reduced to 71.53 g, which accounts for about a 30 % decrease. The results show how the recipe recommendation system contributes to reducing household-generated food waste.