I've been dealing with image labeling recently, having numerous samples that continually increase. Each sample undergoes analysis by an AI, returning around ten labels most likely related to the image (e.g., ['apple', 'car', 'tree', ...]). I convert these labels into corresponding category indices (e.g., [2533, 23, 53, 13, ...]). Attempting to apply Fuzzy C-Means to these labels yielded poor results, likely due to significant index differences. I'm seeking alternative methods to categorize these ten labels into distinct groups using unsupervised learning. With a large number of labels, I aim to output several categories. My goal is to enable users to see other similar images related to a specific label when viewing an image. Any suggestions on achieving this?
The user attempted to apply Fuzzy C-Means clustering to categorize image labels generated by an AI. However, the results were unsatisfactory, possibly due to significant index differences. The user is seeking alternative methods for unsupervised categorization, aiming to output several categories. The expectation is to find an effective way to group the ten labels per image into distinct clusters, allowing users to retrieve similar images based on specific labels when viewing an image.