Cluster Analysis
After assembling the dataset containing fashion-related objects, the subsequent phase involves employing clustering algorithms to discern fashion trends. This process encompasses several key steps:
- Object Embedding: Both the entire image and each detected object are transformed into vectors using a powerful open-source image embedding model.
- Clustering: Advanced clustering algorithms are employed on these object embeddings to pinpoint clusters of objects that share similar concepts.
- Cluster Review and Selection: Clusters that reveal novel concepts are chosen as indicators of emerging fashion trends, making them prime candidates for further analysis and display.