The fast growing trend of electronic waste (e-waste), especially from printed circuit boards (PCBs), has created significant challenges for the efficient recovering of critical raw materials (CRMs). The recovery of valuable materials from these components is crucial for sustaining a circular economy. This paper proposes a adaptive multi-stage transfer learning strategy to improve the automatic detection and localization of selective electronic components on WPCBs containing different distributions of CRMs. We propose an adaptive multi-stage transfer learning strategy using YOLOv11 model that not only improves the detection accuracy, but also addresses the challenges of class imbalance in the WPCBs dataset. The proposed approach improves the detection accuracy by focusing selective electronic components use for CRMs recovery in production processes. This automatic detection and localization system reduces the labor costs, minimizes waste, and supports the principles of sustainability and resource efficiency. Thus, this work presents an effective solution for automated e-waste recycling, contributing to both environmental sustainability and economic growth.