Computer vision’s purpose is to extract meaningful, contextual information from images often in real time. Vision solutions for process automation generally involve a mix of standard camera devices, software, robots and sensors that cannot adapt to changes in context without expert intervention.

At BitMetrics we integrate:

  • Computer vision capabilities such as Deep Learning expertise and traditional machine vision for pattern recognition from images
  • Mathematical Optimization and Machine Learning for robotic task programming (generally grasping)
  • Proprietary databases (togehter with publicly available ones) that facilitate transfer learning for new projects

We specialize in computer vision solutions for dynamic work environments related to (random) bin picking, (random) grasping and classification in quality inspection processes.

Vision algorithms

Classification

Deep Learning based classification algorithms suitable for processes or tasks where:

  • Items or features of one or more categories need to be indentified

  • There is no need to specify the exact location of such features

Location

Deep Learning based location / identificatin algorithms based on regression suitable for visual tasks where:

  • One or several items or features need to be identified

  • Exact location of such item or feature is required

  • Other similar items (i.e., of the same or similar category) may not be of interest and thus need to be ignored

Detection

Deep Learning based object detection algorithms suitable for applications where:

  • Items or features of one or several categories need to be identified

  • Exact location (by means bounding boxes) or segmentation of such items is required