Fashion Trend Analyzer

The Fashion Trend Analyzer is designed to identify and record seasonal costume patterns among different age groups and genders.

1. People Detection and Tracking

• Detection Model: YOLOv8 segmentation model.
• Tracking Algorithm: ByteTrack Algorithm.
• Count: Records the number of people entering (IN) and leaving (OUT).

When a person is entering:

• Segmentation Mask Retrieval: The system fetches the segmentation mask.
• Body Division:
  • The detected person image is divided into three sections based on predefined ratios:
    1. Head Part
    2. Upper Costume Part
    3. Lower Costume Part

2. Age and Gender Prediction (Head Region)

• Models:
  • Uses two separate models for age and gender prediction.
  • Due to real-time application needs, a lightweight model ("coffee model") is employed.
  • For enhanced speed, the model is quantized to INT8.
• Age Categories:
  • The age model can predict the following age categories: '(0-2)', '(4-6)', '(8-12)', '(15-20)', '(25-32)', '(38-43)', '(48-53)', '(60-100)'.
• Gender Prediction:
  • The gender model can predict whether the detected person is Male or Female.

3. Costume Color Detection (Upper and Lower Costume Regions)

• Algorithm: Uses KMeans algorithm-based method for identify colours.
• Steps:
  1. Image Conversion: Converts from BGR to RGB using OpenCV.
  2. Preprocessing:The black background pixels in the pixel mask are removed.
  3. K-Means Clustering: Groups similar pixels (colors) into clusters.
  4. Cluster Analysis: Retrieves the cluster centers representing the most common colors.
  5. Pixel Count: Calculates the number of pixels in each cluster.
  6. Percentage Calculation: Determines the percentage of each color cluster in the image.
  7. Most Common Color Values: Generates a list of common RGB color values along with their respective percentages.
  8. RGB2HSV: Converts the RGB color values to the HSV color format.
  9. Determine Respective Color: Based on the HSV values, the respective color is determined.

4. Data Storage

• Excel: All the records are saved in an Excel file.
• Database: Records are also stored in a MySQL database for reference.

5. Display

• The details, including segmentation masks, are displayed in an OpenCV window.

Demonstration

Video 1

Video 2