Deduplicated benchmark for smartphone conjunctival anaemia detection in children (CP-AnemiC): code, duplicate map and results

Code, duplicate map (meta_groups.csv) and out-of-fold predictions for the article Data leakage inflates smartphone conjunctival pallor anaemia detection in children: a deduplicated benchmark of colour features versus lightweight CNNs on CP-AnemiC. Includes the SHA-256 + perceptual-hash duplicate aud...

Descripción completa

Guardado en:
Autores principales: Fran Franklin Calizaya Apaza, Bisleyn de la Flor Arce Apaza, Jhoel Hairton Apaza Quispe, Ángel Rosendo Condori-Coaquira
Formato: Software
Lenguaje:inglés
Publicado: 2026
Materias:
conjunctival pallor
anemia detection
children under five
smartphone-based screening
data leakage
color features
lightweight convolutional neural network
model latency
explainability
SHAP
Grad-CAM
benchmark
Puno, Peru
Digital Imaging for Blood Diseases
Artificial Intelligence in Healthcare
Retinal Imaging and Analysis
Acceso en línea:Enlace permanente (DOI o Handle)
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Sumario:Code, duplicate map (meta_groups.csv) and out-of-fold predictions for the article Data leakage inflates smartphone conjunctival pallor anaemia detection in children: a deduplicated benchmark of colour features versus lightweight CNNs on CP-AnemiC. Includes the SHA-256 + perceptual-hash duplicate audit of CP-AnemiC (710 rows, 472 unique images), colour-feature classifiers with nested repeated cross-validation, lightweight CNNs (PyTorch/CUDA and Keras) with ONNX/TFLite FP32/FP16/INT8 export and latency measurements, SHAP and Grad-CAM explainability, WHO-2024 threshold sensitivity analysis, per-seed CNN results, figures and a CLAIM 2024 / TRIPOD+AI reporting map. Development phase of a system intended for children aged 6-59 months in Puno, Peru.

Ejemplares similares