Classification, Categorization, Data Cleaning for Machine Learning
How it Works
Tagging image-level tags allows to categorize images, use binary classifier and clean data.
Challenges and solutions:
* Designing meta-data collection for project may include: binary categorization (good/bad image), descriptive value (comments), and class categorization if images are properly cropped or consistent in zoom/size.
* In the Geospatial applications, Image Classification works well for the standard formats such as tile service, allowing sizing up entire imagery data set into standard slices, or tiles, which are further classified with binary or class options. For example, does this tile include a building or not?
Although, without a chance to define borders of the desired object, as it is done in the object detection machine learning models, pure classifiers may lead to difficulties in accuracy, and often get mis-interpreted.
At TaQadam, we used options of annotating data set in different pixel size images compiled of tiles to achieve accurate results.