Data Labeling And AI training – UAV Weed Detection Project
EncordImageClassification
Designed and implemented a data ontology for multiclass classification of weed species in UAV-captured imagery. Labeled large sets of geospatial images to identify invasive weed classes and curated datasets for model training and validation. Increased efficiency and reduced dataset size through binary filtering using early-stage detection models, supporting downstream classification and stakeholder decision-making. • Managed multiclass labeling and project data workflows. • Performed QA and evaluation on labeled imagery. • Collaborated on performance tuning via cloud-based sweep jobs. • Utilized Encord, Roboflow, and Azure ML tools for data labeling and oversight.
Designed and implemented a data ontology for multiclass classification of weed species in UAV-captured imagery. Labeled large sets of geospatial images to identify invasive weed classes and curated datasets for model training and validation. Increased efficiency and reduced dataset size through binary filtering using early-stage detection models, supporting downstream classification and stakeholder decision-making. • Managed multiclass labeling and project data workflows. • Performed QA and evaluation on labeled imagery. • Collaborated on performance tuning via cloud-based sweep jobs. • Utilized Encord, Roboflow, and Azure ML tools for data labeling and oversight.
2025 - 2025