sayon-mitra024/blood-agglutination-classifier

Blood Agglutination Classifier

Confidence-aware calibration framework for medical image classification.

Model Description

This model classifies blood agglutination images as "clumped" or "not clumped" using MobileNetV2 embeddings with AutoGluon Tabular and isotonic calibration.

Usage

from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="sayon-mitra024/blood-agglutination-classifier",
    filename="mobilenet_v2_calibrated/predictor.pkl"
)

Performance

Metric	Value	95%% CI
Macro-F1	0.741	[0.691, 0.789]
Balanced Accuracy	0.766	[0.718, 0.812]