Predicting Raisin Categories using Computer Vision Data

Agricultural Produce Classification

Classifying agricultural products based on morphological features is a key step in automated quality control. In a TechGIG hackathon, the task was to classify raisins into two categories (Keci and Bes) based on 7 morphological features extracted from images.

Algorithm Selection

After testing Logistic Regression, CatBoost, XGBoost, and LightGBM, I constructed a sophisticated ensemble technique that leveraged the strengths of each model to handle the slight overlapping of features between the two raisin varieties.

Results

The ensemble model achieved high accuracy, securing the 14th rank out of 292 participants. This approach proves highly effective for tabular data derived from computer vision feature extraction.