Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{185999,
        author = {Varun Shanmugam},
        title = {Enhancing Fashion Outfit Evaluation through Object Detection, Deep Learning, and Generative AI: Towards Context - Aware Recommendations},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {5},
        pages = {3857-3872},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=185999},
        abstract = {Artificial intelligence is increasingly being utilized in the fashion world. Such applications either dwell on identifying an article of clothing or making dull recommendations because they lack knowledge of the socially contextualized world where these articles of clothing are being used. Therefore, this paper sets up a pipeline approach that integrates object detection with generative ability to develop occasion-driven clothing evaluations. With a trained object detection model, from different fashion-related databases and post-edited with non-maximum suppression, a method detects pieces of clothing from input images pruning redundancies or spurious positives producing judgments. The detection is primarily from the item that the person is wearing compared to the phenotype attributes it may possess, like shirt- or pant-type sleeves, pockets, etc. This extraction then serves as the prompt for an epochal answer from a large language model to remain related to what was extracted and not by hallucination of other articles.
In order to try the pipeline, numerous example images with varying clothing and gender orientations were tested scenarios based on different social contexts and the results were evaluated for detection, non-redundancy, and producing appropriate suggestions. The pipeline successfully functioned; it is possible to evaluate and then produce explainable, situationally-relevant suggestions that are more human-like in style than the early detection generating capabilities. While formative, the benefits and drawbacks of depending on pre-defined visual skeletons and generative methods for fashion use were uncovered. Constraints were due to dataset biases to potential in measuring "appropriate" recommendations. The value is that it bridges the gap between low-level recognition and high-level contextual knowledge through future prospects of generalizable, socioculturally apt, and realistic AI orientation direction fashion recommendation systems.},
        keywords = {},
        month = {October},
        }