Computer Vision: the two questions for five success stories
The sense of sight has played a vital role in the adaptation and survival of species. In fact, nature has solved this challenge from different evolutionary branches, reaching very similar solutions. For example, the morphology of the eyes has been adapted in each species to its environment of survival. Thus, for felines the pupil is vertical, for herbivores it is more horizontal and for mice or humans it is rather circular.
And as in the animal world, companies can also develop their sense of sight, through computer vision or artificial vision, adapting their capacity to different needs to compete and survive . For each problem, different processes can be designed with their own requirements, such as the resolution of the image, which can be low to take a temperature or very high if we are looking to diagnose a pathology.
How does artificial vision generate value?
The artificial vision systematizes information available in
the images, taking different techniques to characterize and interpret its
contents. The main techniques or algorithms are:
• Object detection (star formations)
• Object Classification (Label)
• Segmentation within the image to which
object class each pixel belongs.
By characterizing the content of the images, we can then obtain derived metrics such as the number of objects of a certain class, such as the number of trees on a farm or the number of vehicles (and their brands) on a street. By being able to position the objects and their disposition over time, we can have other derived metrics such as the speed of the object, its trajectory or, in the case of people, it allows us to identify the pose and characterize their reaction to a stimulus.
The use cases in which artificial vision is adding value are multiple, from cases applied to science, such as the monitoring of in vitro fertilized embryos, to other more commercial applications such as the number of people who pass in front of an establishment. The sectors are also multiple, finding applications in sectors as diverse as sports, insurance, health, e-commerce , agriculture or construction. And the functions that it impacts are extensive, all those that use images.
How to ensure the success of a Computer Vision project?
To successfully implement a Computer Vision project, it is
important to clearly identify the use case, what benefit the organization
expects to obtain and from what process. Our experience has shown us that these
projects are multi-user, multi-role and with multiple technological aspects
that add complexity to their management. At Enzyme we have developed a methodology
and the eSTAR.ai platform to ensure that this type of project comes to
fruition. Its main steps are:
• Definition of the use case, target
users and business case
• Launch of project generation of data
set and labeling
• Algorithm training for the business
objective
• Putting into production, monitoring the
quality of your results
The benefits that our methodology provides when implementing these Artificial Vision projects in companies are clear, facilitating customization, improving the time to market of projects, saving time and ensuring the capitalization of project development by all organization teams.
Artificial Vision applied in five success stories
Case 1. Diagnosis of pathology . Applied to the detection of
respiratory diseases in pig farms. It allows increasing the number of animals
evaluated to determine preventive actions such as treatment and selection of
specimens at a lower cost and to maximize the efficiency of the use of vaccines
on farms. Enzyme has been recognized with the IBM Beacon Award of 2021 in AI
and innovation of this solution.
Case 2. Exhibitor or linear analysis . To ensure compliance
with the agreed service levels on how and where to display products on the
distributor's shelf and obtain the maximum amount of information from in-store
displays to correlate with sales volume.
Case 3. School fraud . Facial recognition and verification
of the identity of the students who take the exams remotely.
Case 4. Quality control . from the appearance of the product or packaging. Production standards are set and the different rules and metrics that determine the quality of the product are visually verified. Quality control best example is TC Bolts
Case 5. Medicine . Identify the morphological
characteristics of a tissue to guide the doctor on the areas with the highest
probability of pathological formations.
Artificial vision is a key resource that companies have to
improve their operations and the relationship with their customers. By having
the appropriate methodology, we can make it easier for companies to implement
Computer Vision projects that generate clear profitability through more income
or lower costs for companies.