Fashion Sellability Prediction via Visual and Non-Visual Attribute Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sales prediction techniques for fashion products on e-commerce platforms are inefficient and inaccurate, failing to account for intrinsic visual attributes and extrinsic non-visual factors, leading to challenges in managing inventory effectively.
Innovation Solution
A system and method that combines deep learning models and non-visual parameter models to estimate sellability confidence values based on visual and non-visual attributes of fashion products, using visual attributes determination, non-visual attributes determination, and sellability confidence value estimation to generate an aggregate sellability confidence value.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If historical sales data-based prediction techniques are used, then predictions can be made using available data, but the predictions are time-consuming and inaccurate
Solution Approach 1:
The prediction system segments the analysis into two distinct models: a deep learning model that processes visual attributes (color, fabric, aesthetics) and a non-visual parameter model that processes extrinsic factors (brand, price). This segmentation allows parallel processing of different data types, improving both accuracy and speed while avoiding the time-consuming limitation of traditional single-model approaches.
Solution Approach 2:
The patent introduces visual attribute extraction as an intermediary step between image input and sellability prediction. The deep learning model extracts meaningful visual features (color, fabric, aesthetics) which then serve as inputs to the prediction model. This intermediary processing enables accurate predictions without requiring manual analysis of all historical data, thus reducing time while maintaining precision.
2Measurement precision
If prediction techniques based only on merchandising values are used, then the prediction process is simple, but intrinsic visual factors and extrinsic factors are not considered
Solution Approach 1:
The system segments the prediction task into two specialized models: one handling visual attributes (color, fabric, aesthetics) and another handling non-visual attributes (brand, price, merchandising values). Each model is optimized for its specific data type, enabling comprehensive factor consideration while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The prediction system is designed to handle multiple types of inputs universally - both visual image data and non-visual parameter data. The aggregated sellability confidence value integrates results from both models, creating a universal prediction mechanism that considers all relevant factors (intrinsic visual and extrinsic non-visual) through a unified output metric.
3Measurement precision
If comprehensive analysis of visual and non-visual attributes is performed, then accurate sellability prediction is achieved, but the system complexity increases
Solution Approach 1:
The system divides the comprehensive analysis task into two independent but parallel models, each specializing in one attribute type. This segmentation allows each model to remain relatively simple while the combination provides comprehensive coverage. The modular structure manages complexity by avoiding a single monolithic complex system.
Solution Approach 2:
The patent merges the outputs of the deep learning visual attribute model and the non-visual parameter model into an aggregated sellability confidence value. This combining approach integrates comprehensive information from both visual and non-visual attributes while presenting a unified, manageable output that balances accuracy with system usability.
Data Source
AI summary
A system and method for predicting sellability of a fashion product is provided. The system includes a memory having computer-readable instructions stored therein. The system further includes a processor configured to access one or more catalogue images of a fashion product. The processor is configured to determine a plurality of visual attributes of the fashion product. The processor is further configured to determine a plurality of non-visual attributes corresponding to the fashion product. In addition, the processor is configured to estimate a first sellability confidence value for the reference fashion style using a deep learning model. The first sellability confidence value is estimated based upon the visual attributes. Further, the processor is configured to estimate a second sellability confidence value for the reference fashion style using a non-visual parameter model. The second sellability confidence value is estimated based upon the non-visual attributes. Moreover, the processor is configured to combine the first and second sellability confidence values to determine an aggregate sellability confidence value associated with the reference fashion style.