When applying machine learning, the dependence on data and labelling for creating a reliable Computer Vision algorithm is a challenge. Deep Learning extracts the optimal Algorithm based on the used data, thus we are directly dependent on sufficient data quality.

If implementing Computer Vision software on smaller hardware solutions (edge computing) or if there’s a requirement for greater accuracy or speed than normal, we must optimize the data. Read more about Computer Vision, Image Classification and Object Detection here.

Labelling in Classification : Classifying the picture in one specific category

  1. Data and labels are given – You provide the sorted data and ensure the data has been labelled. (*No Price Guarantee)
  2. We label and sort data – You provide unsorted data, which we then sort and label. (*Price Guarantee)

Labelling in Object Detection: Marking of each object on a picture using polygons, focusing on a high degree of precision. We use the program LabelMe.

When we sort the pictures, we focus on creating, based on our knowledge and experience, a diverse dataset, that in the most optimal way contributes to the desired performance. We boost the performance of our machine learning algorithm by optimizing the data- and labelling-process. With performance in focus the Computer Vision application will solve the problem based on optimal feature extraction capabilities.