AI Tools.

Search
by eugenesiow

edsr-base

edsr-base implements the Enhanced Deep Super-Resolution (EDSR) baseline architecture for single-image super-resolution, trained and evaluated on standard SR benchmarks including DIV2K, Set5, Set14, BSD100, and Urban100. It is part of the super-image library ecosystem and follows the methodology from arxiv:1707.02921. The model upscales low-resolution images without requiring batch normalization layers, which EDSR's authors showed improves SR quality.

Summary text generated by an automated pipeline from the model card · Not individually reviewed or run by us · How this page is made

Share: X Reddit HN LinkedIn

From the model card

Fields below are copied from the tags and counters on the HuggingFace repository eugenesiow/edsr-base at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.

Publisher (HF namespace)
eugenesiow
Library
Transformers
License tag
apache-2.0 — read the license file in the repo before relying on it
Papers cited
arXiv:1707.02921, arXiv:2104.07566
Datasets declared
eugenesiow/Div2k, eugenesiow/Set5, eugenesiow/Set14, eugenesiow/BSD100, eugenesiow/Urban100
Downloads (HF counter at last fetch)
749,992
Likes (HF counter at last fetch)
14
Model card
https://huggingface.co/eugenesiow/edsr-base

Use cases

  • Upscaling low-resolution photographs for print or display
  • Preprocessing satellite or medical images before downstream analysis
  • Benchmarking image super-resolution on standard SR test sets
  • Research reproducing EDSR baseline results on DIV2K or BSD100

Pros

  • Trained and evaluated on five well-established SR benchmark datasets, providing reproducible baselines
  • Apache 2.0 license allows unrestricted commercial and research use
  • Removes batch normalization as per the original EDSR design, which avoids normalization artifacts
  • Integrates with the super-image library for straightforward Python inference
  • Arxiv references link to both the original EDSR paper and a follow-up study for methodological transparency

Cons

  • Base variant is smaller than EDSR-large, trading reconstruction quality for inference speed
  • No pipeline_tag assigned, so auto-discovery in standard Transformers pipelines is not supported
  • Performance on out-of-distribution images (medical, satellite) may degrade versus domain-specific SR models
  • The super-image library has limited adoption compared to mainstream SR frameworks like BasicSR
  • No GPU-optimized quantized variant; large image inference can be slow on CPU

Tags

transformersEDSRsuper-imageimage-super-resolutiondataset:eugenesiow/Div2kdataset:eugenesiow/Set5dataset:eugenesiow/Set14dataset:eugenesiow/BSD100dataset:eugenesiow/Urban100arxiv:1707.02921arxiv:2104.07566license:apache-2.0endpoints_compatibleregion:us