Many tasks that require a big workforce needs to be automated. The consumption of utilities such as electricity, gas, and water is monitored by meters that need to be read by humans, is one of them. The process of traditional water meter reading involves the deployment of trained personnel by the organization to survey every single house and other properties in their list to note down consumption metrics. This data is later aggregated manually and bills are prepared to be sent to the respective consumers for further perusal. A handful of limitations and potential drawbacks to this existing process have a direct impact on cost borne by the organization and level of inconvenience experienced by the consumer. Now the newer technologies in the domain of Computer Vision can be leveraged to make the process cost-effective and streamlined.
Goal of the Project: The goal of the project is to read the exact consumption of water in kilo liters from the image of a water meter using computer vision. In this project, we have designed a three-stage approach for AMR. We first detect the counter region, crop it and then tackle the digit segmentation and recognition stages jointly by leveraging the high capability of Deep Learning techniques i.e. Convolutional Neural Networks. To achieve this goal, the YOLO system (you look only once) has been adopted. The algorithm makes use of a convolutional neural network with special characteristics. To train the network we need to design the image labels, crop it, create the segments containing digits and then read the digits. So below are three stages in which we have divided the project: Detecting the Region of Interest Cropping it Digit Labelling and Recognition
Details of the project is explained in the ppt uploaded. For the data i.e. images of the water meters and all the other related files please follow below links:
Yolov4 Model-1 : https://drive.google.com/drive/folders/1Zuy4ql6NEiIC5KcFV5g5R6AV0VilcoTA?usp=sharing
Yolov4 Model-2 : https://drive.google.com/drive/folders/1uPnmPPHgL111hCNyOMGHtPN9dbdMFeLW?usp=sharing
Thank You!