BharatBreed AI: A Confidence-Aware, Edge-Deployed Image Classification Pipeline for Indian Cattle and Buffalo Breed Recognition

This paper presents BharatBreed AI, a confidence-aware image classification pipeline for selected Indian cattle and buffalo breeds. The system combines EfficientNetB0 transfer learning, image augmentation, TensorFlow Lite conversion, dynamic-range quantization, human verification, and feedback logging. The current prototype uses 360 synthetically generated images across six classes and achieves 91.67% accuracy on a held-out test set of 36 images. The quantized TensorFlow Lite model is approximately 4.6 MB, compared with 16.3 MB for the float32 model. These results demonstrate the engineering feasibility of the proposed pipeline on the development setup; they do not establish real-world breed-recognition accuracy. Evaluation using real, expert-labelled livestock photographs remains future work.