Portrait Generation via Face-Cropping Adversarial Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current picture generation methods require users to manually select and switch between hairstyles, leading to cumbersome operations and poor user experience, with inefficient resource utilization due to multiple invalid operations.
Innovation Solution
A picture generation method using an adversarial neural network model trained through machine learning, which crops the source portrait to isolate the face region and generates a target portrait with a matching hairstyle, eliminating the need for manual selection and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If users manually select and switch between hairstyles, then hairstyle selection flexibility is improved, but operation complexity increases and user experience deteriorates
Solution Approach 1:
The system performs automatic hairstyle selection and picture generation without requiring user interaction. The server autonomously processes the source picture, selects appropriate hairstyles, and generates target pictures, eliminating the need for users to manually switch between hairstyle options.
Solution Approach 2:
The manual mechanical operation of selecting and switching hairstyles is replaced by an automated image processing system using deep learning models. The system substitutes human interaction with algorithmic processing that automatically generates hairstyle-modified pictures.
2Manufacturing precision
If users perform multiple operations to select hairstyles, then hairstyle matching accuracy may be improved, but time consumption increases and productivity decreases
Solution Approach 1:
The system pre-processes the source picture by extracting the face region and preparing it for hairstyle generation before the user requests any action. The deep learning model is pre-trained with extensive hairstyle data, enabling immediate accurate generation without requiring multiple user trials.
Solution Approach 2:
Multiple manual operations for hairstyle selection are replaced by a single automated deep learning inference process. The system substitutes iterative user actions with one-pass algorithmic processing that achieves both high accuracy and efficiency.
3Adaptability or versatility
If the system responds to multiple user operations, then interaction flexibility is improved, but system resource utilization deteriorates
Solution Approach 1:
The server autonomously initiates and completes the entire picture generation process without waiting for or responding to multiple user operations. The system self-manages resource allocation and processing, eliminating wasted computational cycles from handling redundant user interactions.
4Ease of operation
If manual hairstyle selection is implemented, then user control is improved, but device complexity increases
Solution Approach 1:
The complex hairstyle selection and matching logic is extracted from the user interface and relocated to the server-side deep learning system. The terminal device is simplified to only handle basic picture input/output, while the complex processing is performed externally by the server.
Data Source
AI summary
This disclosure relates to a picture generation method and device, a storage medium, and an electronic device. The method includes: obtaining a source portrait picture displaying a target object; cropping the source portrait picture to obtain a face region picture corresponding to a face of the target object excluding a hair portion; inputting the face region picture to a picture generation model to obtain an output result of the picture generation model, the picture generation model being obtained after machine learning training through an adversarial neural network model by using a plurality of sample pictures; and generating a target portrait picture by using the output result of the picture generation model, the target portrait picture displaying a target hairstyle matching the face of the target object. This disclosure resolves the technical problem that pictures generated in related art cannot achieve an effect expected by a user and other technical problems.