SFMformer achieves SOTA image super-resolution with novel Transformer design

Researchers have developed SFMformer, a new lightweight Transformer model for image super-resolution that achieves state-of-the-art results. The model utilizes a novel spatial-frequency modulation approach, combining spatial enhancement with wavelet-domain modulation to improve attention mechanisms. This synergistic design allows SFMformer to outperform traditional additive gains, particularly when modules address different constraints. The model maintains a small parameter count, making it practical for deployment on resource-constrained devices like the Raspberry Pi 5, and has demonstrated top performance across multiple benchmarks. AI

IMPACT Sets new SOTA on image super-resolution benchmarks, demonstrating efficient Transformer architectures for resource-constrained devices.

RANK_REASON The item is a research paper introducing a new model and its performance. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Chih-Hsiang Yang
• peak signal-to-noise ratio
• Raspberry Pi 5
• SFMformer
• Structural Similarity Index Measure
• Transformer

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →