Image recognition with AWS recognition
Gulf Organizational Challenges and Objective
We take care of recognizing license plates, brands, colors and guys of vehicles for information. We automate the process to obtain data in real time, and we connect to a Data Lake to generate reports with valuable information.
How do we help Gulf?
A Batch mechanism was implemented that runs automatically from time to time: it reads the images from Bucket S3, stores images from the cameras of its stations to be processed in batches and thus uses a pre-trained model to recognize the corresponding labels, storing the meta data in a database which will be prepared to be integrated into the Gulf Data Lake for analysis.
1. Training the model in AWS Rekognition:Image recognition model training in Amazon Rekognition to identify information from Gulf gas station images.
2. Image upload automation:Development of a Batch mechanism that reads images from Bucket S3. Cameras store images from their stations to be processed in batches. A pre-trained model is used to recognize labels, storing the meta data in a DB that is ready to be integrated into the Gulf Data Lake for analysis.
3. Quicksight reporting:Generation of reports in Quicksight to establish metrics with valuable information for Gulf. This helps to quantify and evaluate aspects of the business, trends, behaviors and results, to evaluate the performance of the actions and strategies to be implemented.
The model was trained to detect the following information:
- vehicle color
- Vehicle brand
- vehicle type
- Date, time and camera IDs
- dispenser position
- vehicle plate
Technologies implemented in the development of the project
A model for recognizing specific information in images was developed with the help of different AWS services. Which works in the following way, the cameras of the Gulf branches are connected to an S3 in which the images are stored and pass through the Rekognition service and store the recognized information in a DynamoDB database.
This process is automated through AWS Batch, to execute it from time to time. Once having the source of the model, it is connected to the Data Lake through Lake Formation and Athena to later generate the Quicksight reports.
Advantages and improvements of the project
Using AWS technologies made it possible to build a sufficiently robust and trained model to be able to detect characteristics within the images, such as: license plate, date, time, color, brand, position and type of vehicle of its clients.
The Rekognition model at the business level helped the client to get to know their clients better, since all the information is collected within the model in order to obtain a detailed report. This report is used to analyze consumption and long-term prospects.
Challenges and objectives that were achieved in the development of the Project
- A Machine Learning model fed by images identified by the system was created.
- Brand recognition and vehicle logos.
- Recognition of primary colors in an image taking light and shadow into account.
- Recognition of types of vehicles according to their specific characteristics.
- An efficient processing of the Machine Learning model was achieved to recognize a high volume of images.
Get the power of digital solutions in your company
Companies around the world have trusted us. We want you to be part of this!
Codster works hand in hand with the SNTSS Provident and Savings Fund to develop a mobile application where users can check savings balances, request withholdings, schedule appointments and more financial services through a Mobile Application.Know our work
We integrate the sales data into the Drink Fitzer platforms. We integrate state-of-the-art technologies such as MiPos, Bind, Shopify, Deliverect and Amazon API Gateway, which allows control of orders, inventories, points of sale and assortments in stores.Know our work
We developed a digital file system that can be replicated in various clients, where your collaborators can upload files to generate a file per client and within the system, be able to monitor the credit approval process.Know our work
Codster makes it possible for Gulf to integrate image recognition technologies using Amazon Rekognition to automate the identification of images in service stations in order to obtain a detailed report of its customers in real time.Know our work
Codster partners with Talentum HR Company to develop an interactive Website, and design a UI/UX design strategy, to offer the community recruitment services in regions such as North America, Central America, South America and the Caribbean.Know our work
Codster works together with Pando to offer the community a digital platform or Marketplace where millions of citizens can exercise their social responsibilities. Pando offers people the possibility of active and digital political participation.Know our work
Codster and Prometec develop and design a technology strategy for Prometec products, with the aim of implementing cloud services with Azure and a DevOps strategy, we focus on creating improvement plans at the architecture level for this company.Know our work
Codster, together with BpB, designed a system that fits into the current process of being able to take the request to a digital appraisal, which extracts the initial information from the client and performs two automatic processes through APIs from a third party to obtain urbanization data, marketing to through AWS Lambda.Know our work