IMAGINE
Envision
CompSoc
Abstract
The project focuses on converting rough and noisy hand-drawn sketches into realistic and meaningful images.
An autoencoder model is used to enhance and denoise the input sketch before image generation.
A pretrained Stable Diffusion model then generates high-quality images guided by both the refined sketch and a user-provided text prompt.
The system also includes an image captioning module to generate a textual description of the final output image.
Semantic similarity between the generated caption and the original prompt is calculated to evaluate the quality and relevance of the output.
Overall, the project aims to combine computer vision, deep learning, and natural language processing into a complete multimodal AI pipeline.
Abstract
The project focuses on converting rough and noisy hand-drawn sketches into realistic and meaningful images.
An autoencoder model is used to enhance and denoise the input sketch before image generation.
A pretrained Stable Diffusion model then generates high-quality images guided by both the refined sketch and a user-provided text prompt.
The system also includes an image captioning module to generate a textual description of the final output image.
Semantic similarity between the generated caption and the original prompt is calculated to evaluate the quality and relevance of the output.
Overall, the project aims to combine computer vision, deep learning, and natural language processing into a complete multimodal AI pipeline.
Report Information
Team Members
Team Members
Report Details
Created: May 12, 2026, 3:10 p.m.
Approved by: None
Approval date: None
Report Details
Created: May 12, 2026, 3:10 p.m.
Approved by: None
Approval date: None