Kadadevarmath , J. CATR-Net: A Cross-Attention-Driven Transformer–CNN Framework With Bidirectional Feature Fusion for High-Accuracy Image Recognition. IJACECT 2026, 15, 207-220.