Vision Based Currency Counting for An Automated Donation Box

India is a country renowned as the land of temples serving as a center of spirituality, where people believe in the divine and offer donations as money, jewels, etc. based on their economic status. It is known as ‘kaanike’ and is offered as a symbol of gratitude, devotion, and faith. Almost all temples across the world have a donation box called hundi. These hundis mainly consist of a large number of coins and banknotes. The counting of coins and banknotes requires more human intervention and is a tedious job. Hence, this process of counting coins and banknotes can be automated so that efficiency will be improved and the time consumption of the process will be reduced. In this work we have developed a hundi in which the counting of coins and banknotes will be automated. The system is developed such that the coins and banknotes offered by the devotees will be scanned by the camera, and by using the convolutional neural network (CNN), the denominations of the coins and banknotes are determined. Finally, the cumulative amount is computed by adding the monetary values associated with each coin/banknote.