The application of machine learning and computer vision for the financial services industry can greatly improve the speed, accuracy and efficiency of services to consumers. Combining the use of entity extraction and optical character recognition, large volumes of documents can be classified to streamline the automation of KYC processes or applications for loans and credit. Identifying spending habits of consumers will allow machine learning models to provide more accurate credit scores to better assess the risk profile of consumers and even identify fraud transactions.
Other applications that could be used for financial services are:
Our client, a financial comparison service provider, required an automated method to review large amounts of transactional statements to provide quick responses to their customers for evaluating credit scores and eligibility for various credit applications (e.g. personal loans, credit cards etc.). The increase in speed of response would result in a higher uptake of their services and reduce the resources required to manually review documentation.
Enlabeler collaborated with CapeAI, who was responsible for building the machine learning model that would automate the evaluation of the financial statements. Enlabeler provided the training dataset from 40,000 transactions, the data classification required identifying the transaction type and its source category. A team of 8 labelers with a background in finance or accounting was employed to complete the project in a short time span of 2.5 weeks.
8 domain experts with accounting or economics background
Categorize 40,000+ transactions
Completed in 2.5 weeks
The health care industry can leverage machine learning to better serve their patients by using it to identify and diagnose patients, support medical research and development, run system analyses and so much more. The right data annotation partner can handle medical imagery and videos securely, accurately extract and annotate large amounts of high quality training data for robust ML models – choose Enlabeler. Other examples of use cases in the health-tech space are:
Decoded Health is developing a Clinical Hyper-automation Platform which will allow doctors and physicians to reach and help more patients. They have high volumes of unstructured medical data that require high quality label classification to feed their AI-driven solution. Decoded Health also wanted to partner with an organization that aligns with their values of making a social impact.
Enlabeler’s humans-in-the-loop solution brought together the right people and processes to efficiently create the training data. Medically trained and qualified individuals were sourced to project manage, identify and classify medical text as well as provide quality control.
The partnership between Decoded Health and Enlabeler is an innovative way to improve doctor-patient relationship and create more opportunities in the medical technology field.
5 medical professionals providing classifications
High volumes 14200 classifications completed since March 2022 (on going)
8 successful projects passed
Assisted in helping to cover 90% of patient diagnosis to accelerate their platform deployment
“Enlabeler has been a key partner in helping Decoded Health execute its vision for health care. With quality and responsiveness, Enlabeler has speedily supplied us with a trove of labeled health data that powers our platform’s inferential capabilities, enabling physicians to help more patients and touch more lives.”
Brent K Sugimoto
Chief Medical Officer
Transcription and translation used for subtitling in movies and TV productions helps reach a wider audience. Meaningful stories can be told to an audience in their native language, spreading the message and enjoyment of these media productions. There is also the added benefit of SEO making these videos more searchable if published online. Other examples that the media and entertainment industry can benefit are:
A popular South African telenovela TV show wanted to extend its reach into the African-Portuguese and Swahili speaking market . Viewership from these communities are limited due to lack of professional resources and infrastructure to do subtitling. The client needed a feasible solution for creating a framework to translate and transcribe English into various African languages dialects that connect with its audience
Enlabeler offered a fully managed team of translators and transcribers that specialized in these under-resourced African languages. The team provided translation and transcription in multiple African languages and dialects that was specific to the region the TV show was to be aired. Quality control was essential to ensure that the right dialect was used to engage and tell a compelling story to the viewers.
450 hours of video content was processed
3 native African languages were used across translations and transcriptions
As our energy needs as a species expands so does our need for control, monitoring and reporting on the state of our planet. Cataloging and annotating data from multiple geographical data sources help improve predictions of specialized geological cases. Combined further with human energy consumption data, predictive energy models can optimize our energy efficiency whilst maintaining sustainability of the planet.
Other examples that infrastructure can benefit are:
Our client is on a mission to clean up the ocean by developing scalable technology to make these efforts more efficient. They required a hands-on outsourced team that could be activated timeously to annotate large volumes of images. The data output would assist their machine learning model to locate polluted areas and deploy cleanup solutions efficiently.
Enlabeler assembled a team of 20 annotators that could be easily activated to provide a quick turnaround. The team diligently reviewed 6000 aerial and crowd sourced mobile images of the ocean, coastline and rivers to identify waste. The output of the data had an 89% accuracy rate on the model and allowed the deployment of cleanups at 5 different locations.
Total of 6000 images reviewed
140k of annotations
89% accuracy on the model
20 jobs created
The use of machine learning for infrastructure makes planning, constructing and maintaining the assets more feasible and mitigates potential risk. Annotated datasets from aerial images of existing landmarks to identify physical objects such as trees, electricity poles and transmission lines can aid in automated inspections and monitor safety. Other examples that infrastructure can benefit are:
The future of road infrastructure is moving towards digitalisation for better management and maintenance. Fugro required large volumes of data to be processed with labeling assets on the road networks – something that was not developed in the region yet.
Enlabeler was able to process the high volume in a short time span. A dedicated project manager worked closely with the Fugro and labeling team made sure that the requirements were met to standards and for an efficient delivery. The project was completed with on-premise data and at the best market value rate for a project as specialized as road asset management.
8,300+ images completed in 4 weeks
200+ images processed daily
The use of data labeling for computer vision models in agriculture can increase produce throughput and workflow efficiency for precision growing. Training datasets to accurately recognise produce or identify plant defects can provide prompt yield estimates to assist with farming operations. Other examples that agritech can benefit are:
Our client, a leading agritech company, had ±381,000 acres of aerial data to process. The client required a local partner to output a high turnaround of labeled and annotated aerial imagery. As part of the client’s ethos to assist local farmers and communities, they wanted to collaborate with an organization that has a mission to make an impact on the local community.
With Enlabeler’s goal to expand Africa’s digital talent ecosystem, we accessed untapped talent and provided 40 jobs to youth that were unemployed. The team achieved a 95% turnaround time within 24 hours from the initial intake of aerial data. The team was required to accurately detect objects and segments of the produce and agriculture to determine yield and crop estimation.
95% turnaround time within 24hrs
36k acres processed per month
40 jobs created
An extension of the agritech industry is livestock management where big data can be used to maintain the well being of the animals to optimize growth cycles and resolve day-to-day operational challenges. Supply chain management can also be improved by predicting produce yield and quality for supply and demand forecasting. Other areas where livestock monitoring would be beneficial are:
Serket is a leader in digital solutions for pig farming to assist with the livestock management process. They wanted to predict the behavior of pigs in the pen by creating a model that looked at the actions and posture of each pig. Serket had adopted a new tool to allow them to input the data and collaborate easier. They had hours of surveillance videos they had recorded without any labeled information.
Enlabeler’s agnostic approach means that we can focus on the technical and advanced annotations, regardless of the platform. Our team was trained and set up on the client’s tool and could seamlessly follow the guiding principles that outlined their training model. We were able to produce a quick and accurate turnaround with rapid results from the output.
300+ videos equates to 9,000+ images processed
51 300 pigs annotated
23 unique jobs created
We scaled 75% more for the second iteration
“At Serket we are building a system to recognise and analyse the behavior of pigs. We always receive good quality data from Enlabeler. They are known for high quality adaptability and excellent communication.”