Upload inference.py

inference.py

@@ -1,20 +1,21 @@
 """
-TAS-BB-v1 MEWC-Keras Classification Model - Custom Inference Script
-
-This script provides model-specific inference code for the Tasmania
-species classifier (96 classes). It follows the AddaxAI-WebUI interface contract
-for custom classification models.
-
-Model: Tasmania MEWC Ensemble (tas_ens_mewc.keras)
-Framework: Keras 3 with JAX backend (TensorFlow compatible)
-Classes: 96 classes (Tasmanian terrestrial mammals and birds)
-Training data: 2.5+ million images
-
-MEWC - Mega Efficient Wildlife Classifier - University of Tasmania
-Original source: streamlit-AddaxAI/classification/model_types/mewc-keras/classify_detections.py
-Reference: https://github.com/zaandahl/mewc
-Adapted by: Claude Code on 2026-01-11
-Updated: 2026-01-13 - Migrated to class-based interface
+Inference script for TAS-BB-v1 (Tasmania MEWC Species Classifier)
+
+MEWC (Mega Efficient Wildlife Classifier) for Tasmania trained on 2.5 million labelled
+images from 96 classes. Includes all non-volant terrestrial mammals (native and introduced)
+and 50+ commonly observed bird species. Overall accuracy and F1 scores exceed 99%.
+
+Model: Tasmania MEWC Ensemble
+Input: 224x224 RGB images
+Framework: Keras 3 with JAX backend (EfficientNet v2 Small architecture)
+Classes: 96 Tasmanian terrestrial mammals and birds
+Developer: Barry Brook (University of Tasmania)
+Citation: https://ecoevorxiv.org/repository/view/6405/
+License: CC BY 4.0
+Info: https://github.com/zaandahl/mewc
+
+Author: Peter van Lunteren
+Created: 2026-01-14
 """
 
 from __future__ import annotations