We supported Universidad Santo Tomás with software to recognise and classify fresh-smear microscopy images within an ANID project being carried out by the university.
Conceptual diagrams without sample images or project data. They do not represent findings or performance metrics.
Applied research
Computer vision and machine learning
Software engineering and models
The research context
Connecting a research question with a usable solution.
The project involved fresh-smear microscopy images and supporting their recognition and classification. The challenge covered both modelling the visual task and building software to process images and review results.
We contributed software engineering, modelling, model training and application development, connecting these parts within a single workflow.
Our contribution
Engineering, models and software working together.
We carried the task from definition to an application connecting images, classification and research records.
01
Design the solution before building it
We carried out software engineering to connect image uploads, processing, classification and results review. The design considered the workflow and the relationship between images and their records.
Conceptual schematic01 / 05
02
Model the recognition task
We organised classification into a sequence of tasks: recognise smear images, assign a category and perform a further classification where applicable. This connected the visual question with defined outputs for the project.
Conceptual schematic02 / 05
03
Train models for the visual task
We worked on training classification models from image examples. Specialised models were integrated into a processing chain, rather than treating each part as an isolated tool.
Conceptual schematic03 / 05
04
Bring models into image analysis
We developed image preparation and classification within the software. The application can process sets of images and present estimated categories alongside their results.
Conceptual schematic04 / 05
05
Organise results for later review
We connected images and classifications with dated records and locations. History, summaries and report exports let results be reviewed with their context.
Conceptual schematic05 / 05
What we made possible
A connected workflow for working with images.
01
An integrated solution
Software engineering and classification models brought together in an application.
02
Results for review
Images and estimated categories connected to their records.
03
Organised information
History, summaries and reports to support the project’s work.
We connected modelling and training with software development so classification could become part of a workflow. Interpretation of the results remains with the research team.
The work behind the case
Your next step
Does your research need to connect models and software?
Let’s discuss the task, available images and how to work with the results.