Member Details:
| Name | Student ID | Username |
|---|---|---|
| Priyanka PDMK | IT19954974 | Modeesha-Kalani |
| Jayawardana GVHD | IT19972176 | HansakaDilshanJayawardana |
| Dharmadasa KRWS | IT19970578 | SandunDharmadasa |
| Alahakoon ASGP | IT19960432 | ShalithaAlahakoon |
In different cultures, clothing may be an indication of age, social standing, lifestyle, and gender. People can be recognized by their clothing, for example, "the man in the black coat" or "the woman in the red high heels." Given the significance of clothes in culture, there are several uses for the word fashion classification. Finding the most comparable fashion items in an e-commerce database, for instance, may be made simpler by predicting apparel features from an unlabeled photo. Similar to this, an automated fashion stylist who may suggest outfits based on the user's anticipated style can be powered by the classification of a user's preferred fashion photographs. In surveillance circumstances where information about people's attire might be used to identify criminal suspects, clothing identification in real time may be helpful. The automatic annotation of photographs with tags or descriptions linked to clothes is made easier by fashion classification, improving information retrieval in contexts like social network users' photos.
The most pressing issues to address may differ depending on how fashion classification is used. By annotating photographs and identifying the most comparable fashion items to a fashion item in a query image, we will focus on enhancing fashion classification. Using publicly available data, two powerful deep learning architectures are used to investigate, examine, and compare fashion photograph analysis. It has several columns and records with a variety of photographs. There is a distinct photograph in each row. In order to give an accurate prediction, these models are trained using the actual results. Using publicly available data, four powerful machine learning architectures are used to explore, analyze, and compare fashion photograph analysis. It has several columns and records with a variety of photographs. There is a distinct photograph in each row. In order to give an accurate prediction, these models are trained using the actual results.
The Kaggle platform is where the data set is gathered. Different techniques are used to perform supervised learning architecture models. In this report, the classification of fashion photographs using supervised CNN architecture, ANN architecture, VGG19 architecture, and KNN architecture are calculated, quantified, and analyzed. The effectiveness and result accuracy of these four models are contrasted.
Priyanka PDMK - ANN Architecture using Fashion MNIST Dataset | Jayawardana GVHD - CNN Architecture using Fashion MNIST Dataset | Dharmadasa KRWS - Transfer Learning with VGG19 Architecture using Fashion MNIST Dataset | Alahakoon ASGP - KNN Architecture using Fashion MNIST Dataset
Language - Python | Model Training Tool - Google Colab | Integrate Technology Service - GitHub
- Dataset Link: https://www.kaggle.com/datasets/zalando-research/fashionmnist
- Reporsitory Link: https://github.com/HansakaDilshanJayawardana/MLAssignment
- YouTube Link: