OPTIMIZING THE POTENTIAL OF URBAN FARMING BY USING GOOGLE STREET VIEW AND MACHINE LEARNING

• Main Author: Fauzan Adziima, Andri
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/71790
• Institution: Institut Teknologi Bandung

Google Street View and machine learning can be powerful tools for optimalizing the potential of urban farming, by providing valuable information about site suitability. From this study, it was found that the CNN (Convolutional Neural Network) algorithm, especially restnets, can be used to properly classify building and road types, however with the additional optimization using the SVM (Support Vector Machine) algorithm it can reach even better accuracy to identify suitable locations to develop urban farming area. In the classification process using the CNN (restnet) algorithm, this algorithm can classify with an accuracy of 0.88, while by optimizing with the SVM algorithm the accuracy score can be increased to 0.97 or an increase of 0.09. The estimation results for the potential urban farming area in Bandung City obtained a value of 21,882 m2 – 48,784 m2 (35,333 m2 average) / 2.19 ha – 4.88 ha (3.53 ha average), with an average the area of the urban farming in each house ranges from 5.29 m2 - 11.79 m2 (average 8.54 m2), it is also known that the potential for urban farming areas tends to be spread out in the suburbs of Bandung City, this is because the majority of the housing is spread out in the suburbs, whereas in the city center it tends to contain buildings that are used as commercial areas. It can also be seen that the highest distribution of urban farming areas is in the Lengkong and Astanaanyar sub-districts. For future work, we would like to use our trained CNN algorithm with the addition of a more complete dataset, which is expected to improve the accuracy of Street View image classification. In addition, we also want to develop this research not only to classify urban farming, but on other research topics, because basically Google street view imagery can provide many useful insights for policy makers in managing their area efficiently.