Service for AI-based land cover class identification for LPIS update

Service for automatic identification of land cover classes: Provision and operation of an IT service in a SaaS model for the automatic identification of land cover classes using artificial intelligence (AI) based on geospatial data, for the update of the Land Parcel Identification System (LPIS). This includes the identification of obligatory basic categories: communication areas, forests, industrial and urban areas, and homesteads with buildings. For these, minimum quality indicators of Overall Accuracy 0.85, Recall 0.80, and Precision 0.80 must be met. Additionally, non-obligatory premium categories such as waters and other non-agricultural areas can be offered, which will be considered if they meet the minimum quality indicators of the basic categories. The scope of services includes the processing of raster and vector geospatial data, the generation of vector layers with classification results and confidence levels, ensuring geometric correctness and topological consistency, as well as the visualization, analysis, editing, and export of data in standard formats. Data processing must handle large volumes using parallel processing. The contractor provides the data processing environment, software, AI models, and their mechanisms, enables the selection and import of input data, delivers quality metrics (Overall Accuracy, Precision, Recall), and ensures availability, security, and reliability. Furthermore, technical support, the provision of user and technical documentation, user training, and up to 50 man-hours of consultation in Polish are required. The graphical user interface (GUI) and documentation must be in Polish. Classification results, along with output data and confusion matrices, must be presented in a dedicated workspace.

Tender documents

Additional details