Real-Time Cooked Food Image Synthesis and Visual Cooking Progress Monitoring on Edge Devices

2025-11-21

AI Summary

To address the dual challenges of insufficient realism and excessive computational overhead in raw-to-cooked food image generation on edge devices, this work introduces the first fine-grained dataset tailored to oven-based cooking processes, annotated by professional chefs with maturity labels, and proposes Cooking-aware Image Similarity (CIS), a novel metric jointly evaluating generation fidelity and culinary plausibility. Methodologically, we design a lightweight conditional generative network conditioned on recipe text and real-time cooking state, optimized via CIS-guided loss to ensure temporal coherence and domain authenticity; it further enables user-preference-driven, controllable visualization of doneness progression. Evaluated on both our proprietary and public benchmarks, our approach achieves 30% and 60% FID improvements over prior art, respectively—marking the first demonstration of efficient, realistic, and interpretable cooking image synthesis with synchronized progress perception on resource-constrained edge platforms.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Cognitive Modeling & Cognitive Systems: Computational Creativity
• Machine Learning: Learning on the Edge & Model Compression

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Synthesizing realistic cooked food images from raw inputs on edge devices is a challenging generative task, requiring models to capture complex changes in texture, color and structure during cooking. Existing image-to-image generation methods often produce unrealistic results or are too resource-intensive for edge deployment. We introduce the first oven-based cooking-progression dataset with chef-annotated doneness levels and propose an edge-efficient recipe and cooking state guided generator that synthesizes realistic food images conditioned on raw food image. This formulation enables user-preferred visual targets rather than fixed presets. To ensure temporal consistency and culinary plausibility, we introduce a domain-specific extit{Culinary Image Similarity (CIS)} metric, which serves both as a training loss and a progress-monitoring signal. Our model outperforms existing baselines with significant reductions in FID scores (30% improvement on our dataset; 60% on public datasets)

Problem

• Synthesizing realistic cooked food images on resource-constrained edge devices
• Overcoming unrealistic outputs from existing image generation methods
• Ensuring temporal consistency and culinary plausibility in cooking progression

Innovation

• Edge-efficient generator guided by recipe and cooking state
• Culinary Image Similarity metric for training and monitoring
• Synthesizes realistic food images from raw inputs on devices