PSCount: Dataset and Pre-trained Model for Polystyrene Particle Counting

This dataset accompanies the PSCount software for automated polystyrene particle counting in brightfield microscopy images. Contents:- Training data: 19 labeled images with point annotations and generated masks- Test data: 24 images with ground truth (Set A: point annotations, Set B: CSV)- Pre-trained model: Fine-tuned Cellpose-SAM model- Output: Prediction results and probability maps The fine-tuned model achieves MAE of 2.1 particles per image, representing a 97% reduction in counting error compared to the base Cellpose-SAM model. Code available at: https://github.com/seoleuns/pscount