The visual analysis and interpretation of context is a need across several sectors of activity, particularly for repetitive tasks that normally fall to human operators. Computer vision based on classical approaches or supported by Artificial Intelligence (AI) applied to data with inherently visual representation, allows the automatic extraction of patterns and occurrences of interest, supporting or automating processes that, traditionally, are carried out by human observers in different contexts, from industry to medicine.
The CCG/ZGDV Institute has a team specialized in developing approaches based on image processing and computer vision, applied to address underlying issues of ergonomics and human error, in carrying out active monitoring and detection of patterns and occurrences of interest, in isolation. or in sequence context. In this way, it is possible to free up repetitive tasks from human operators, in examples such as highlighting defects in an industrial manufacturing product or inspecting high-resolution medical images in order to detect a health problem, among other applications.
Activities
CCG/ZGDV supports organizations through the following activities:
- Support in identifying active or passive equipment for image acquisition or scene reconstruction, depending on needs;
- Development of approaches based on image processing and classical computer vision and/or AI for data with inherently visual representation, depending on the challenge to be addressed (classification, object detection, sequence analysis...)
- Survey and selection of the most appropriate techniques and methods to perform visual analysis/interpretation of context, through the challenge
- Implementation of strategies for continuous improvement of inference and processing parallelization models, with a view to promoting system performance
- Integral construction of the logic of innovative vision systems that enhance the digital transition, following good practices underlying optical inspection strategies for contextual inference
Benefits
01
Reduce the rate of errors arising from human observation
Implement resilient computer vision approaches, through image/video analysis, or 3D reconstruction analysis, with the aim of considerably reducing errors caused by human factors and increasing the effective impact on productivity and cost reduction.
02
Enhance the improvement of ergonomic conditions
Apply automatic vision systems in order to free human operators from more ergonomically demanding activities, contributing to improving workplace well-being.
03
Prioritize comprehensive analysis over sampling
Implement computer vision with a focus on low-cadence performance, through analytical processes that extend to the entire spectrum of elements of interest in a given context, to the detriment of sampling, contributing to the reduction of error rates.
04
Support scalability of inference approaches
Allow the learning process of vision approaches to not be stagnant and enable the continuous progression of inference capabilities. To this end, we use human-in-the-loop approaches that enable the reinforcement of computational models on a continuous basis, as well as the teaching of new classes of occurrences.
