E-commerce stands as a continuously expanding sector within the Indian economy, offering businesses the means to effectively serve a broad spectrum of customers, thus enhancing their reach. Fashion holds substantial societal significance, and the fashion industry has undergone humongous growth. The advancement of e-commerce has streamlined the buying and selling of fashion apparel for both sellers and buyers. One challenge faced by sellers is the manual tagging of product images when uploading them to the platform, which can result in misclassifications and cause products to go unnoticed in search results. An appropriate image annotation is a major challenge in the field of fashion e-commerce. Consequently, there arises a need for an object detection model that is capable of automatically identifying and tagging fashion items. However, after analyzing and experimenting on various object detection models on fashion dataset, it becomes apparent that depending solely on a single model is insufficient. To address this limitation of using a single object detection model, the proposed system is designed to combine the predictions of multiple object detection models and results in the most accurate and appropriate results by using an ensemble learning approach. The proposed approach is known as Non-linear Maximal Weighted Box Integration (NWI). This approach has significantly improved the results compared to what individual object detection models can achieve. Also, shown significant results over existing ensemble approaches.