Pith. sign in

REVIEW 3 major objections 5 minor 105 references

Text-Aware Image Restoration with Diffusion Models

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Diffusion-based restoration can be made text-aware by joint training with a text-spotting module that feeds recognized text back into the denoising loop, and the paper shows this beats general-purpose restorers on text recognition accuracy.

desk verdict Defines a useful new task with a strong benchmark and a sensible method; the main caveat is the curation pipeline filters out the hardest text, so real-world gains are likely smaller than the headline numbers. read the letter →

arxiv 2506.09993 v2 pith:BDGTOJEN submitted 2025-06-11 cs.CV cs.AIcs.LG

Jaewon Min , Jin Hyeon Kim , Paul Hyunbin Cho , Jaeeun Lee , Jihye Park , Minkyu Park , Sangpil Kim , Hyunhee Park
show 1 more author
Seungryong Kim
This is my paper
classification cs.CVcs.AIcs.LG
keywords text-awareimagerestorationdiffusionmodelstext-imagehallucinationtextspottingscenerecognitionbenchmarkvision-languagemodelfilteringtextualpromptguidance
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that image restoration can be made to respect the text in a scene instead of hallucinating plausible-looking but wrong characters. It introduces Text-Aware Image Restoration (TAIR), a task that asks for the simultaneous recovery of scene appearance and textual fidelity, and argues that existing diffusion-based restorers fail at this because their generative priors synthesize text-like textures rather than the intended words. To support the task, it builds SA-Text, a 100K-image benchmark of high-resolution scenes with VLM-verified text annotations, and proposes TeReDiff, a diffusion restoration model jointly trained with a text-spotting module so that recognized text guides the denoising process. A sympathetic reader would care because readable text in restored photos matters for documents, street signs, and navigation, and the paper's experiments show its model consistently beats the best general restorers on text-spotting accuracy.

What carries the argument

The two load-bearing components are the SA-Text curation pipeline and the recognition-guided prompt loop. The pipeline runs a text detection model on full-resolution images and again on crops, then keeps a text instance only if two vision-language models transcribe it identically, and finally discards blurry crops using a third VLM judgment; this produces 100K sharp, densely annotated training images. The prompt loop is the mechanism that carries the restoration: at each denoising timestep, the text-spotting module outputs detected polygons and recognized characters, the recognized text is turned into a natural-language prompt, and that prompt conditions the next denoising step, letting intermediate recognition correct the image being generated.

What would settle it

Take a test set of real degraded images with independently verified ground-truth text, deliberately oversampling cases where two strong OCR systems disagree on the low-quality input (the very cases SA-Text's curation discards); if end-to-end text recognition F1 on this hard subset is no better than the baseline, or if supplying the ground-truth text as the prompt does not improve the final image, the paper's central claim that recognition-guided prompting faithfully restores text would be contradicted.

Watch

Extended reading notes

Core claim

TeReDiff is a multi-task latent diffusion framework in which a ControlNet-conditioned U-Net restores the full image while a transformer-based text-spotting module reads from the diffusion decoder's multi-scale features. Training proceeds in stages: first the restoration module is trained with text prompts, then the spotting module learns from diffusion features, then both are optimized jointly. At inference, the spotting module's recognized text is formatted into a prompt ("A realistic scene where the texts ... appear clearly on signs...") and fed into the next denoising step, so characters are progressively corrected rather than regenerated from scratch. On SA-Text and on a real-world test set, the paper reports the best detection and end-to-end recognition F1 scores at every degradation level among the compared GAN and diffusion restoration baselines, and better reference-based image quality than DiffBIR.

Load-bearing premise

The load-bearing premise is that the benchmark's ground-truth text annotations are trustworthy because they come from two vision-language models agreeing on every character; if that filter quietly drops the hardest, most ambiguous, or most corrupted text instances, then the measured recognition gains overstate how well the model restores text in the real world.

Editorial extensions

If this is right

  • Restoration of scenes containing text improves recognizability: off-the-shelf text spotters achieve higher F1 on TeReDiff restorations than on those from any compared baseline, at all tested degradation strengths.
  • Text fidelity and image quality are not in conflict: joint training improves reference-based perceptual metrics (PSNR, SSIM, LPIPS, DISTS, FID) over the DiffBIR baseline.
  • Diffusion features alone are enough to train a text-spotting model from scratch, even with 20K samples, where a ResNet backbone fails to learn recognition.
  • Ground-truth text prompts define an upper bound on restoration accuracy: supplying correct text during denoising yields large gains, so the remaining gap to the bound measures the spotting module's error.
  • The fully automatic curation pipeline is scalable to larger corpora, making the TAIR task extensible beyond 100K images.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The dual-VLM agreement filter defines ground truth by current VLM capability; if a VLM systematically fails on a class of degraded text (e.g., stylized fonts, low contrast), that class is silently removed from SA-Text, so reported accuracy gains may not transfer to harder real-world text.
  • The prompt loop is a closed feedback system: a misread at an early timestep becomes the prompt that steers later denoising, so errors could compound; a testable extension is to weight prompts by recognition confidence or to refresh the prompt only at selected timesteps.
  • The same spot-and-prompt mechanism could apply to other symbol-critical restoration tasks, such as barcodes, QR codes, license plates, or mathematical notation, where pixel fidelity matters less than exact symbolic reproduction.
  • Because SA-Text sources from SA-1B, the benchmark inherits SA-1B's distribution; the paper acknowledges performance declines for small text, but does not quantify how accuracy varies as text-instance size shrinks.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a new task, Text-Aware Image Restoration (TAIR), which requires restoring full natural images while preserving the fidelity of embedded text. To support this task, the authors introduce SA-Text, a 100K-image benchmark automatically curated from SA-1B by running a text detector, re-detecting on crops, and retaining only text instances that two vision-language models (Qwen2.5-VL and OVIS2) transcribe identically and that a third VLM-based blur filter classifies as sharp. They also propose TeReDiff, a diffusion-based restoration model that jointly trains a text-spotting module on diffusion U-Net features and uses the spotted text as a prompt in subsequent denoising steps. Experiments on SA-Text and a newly curated Real-Text set show that TeReDiff achieves higher text detection/recognition F1-scores than existing restoration baselines (e.g., DiffBIR, StableSR, SeeSR, FaithDiff, SUPIR) across three degradation levels, while image-quality metrics are reported only against DiffBIR, with worse no-reference scores on SA-Text.

Significance. If the results hold, the paper fills a genuine gap: general-purpose diffusion restoration methods do hallucinate or blur text, and the proposed joint text-spotting/restoration framework is a sensible way to address this. The SA-Text dataset is large and could become a useful resource for the community, especially the demonstration that diffusion features can train a text spotter with limited data. The ablations (multi-stage training, prompt styles, null vs. predicted vs. ground-truth prompts) are informative and give partial credit to the claimed mechanism. However, the evaluation protocol has a load-bearing weakness: the benchmark and its test sets are built with the same automatic curation pipeline that explicitly discards hard-to-read and blurry text, so the reported gains may not transfer to the difficult real-world cases that motivate the task. The image-quality claims are also under-supported because only one baseline is compared. These issues are addressable with additional experiments and analyses, so the paper merits a major revision rather than rejection.

major comments (3)
  1. [Sec. 3.1, Sec. 5.1, Tab. 2, Tab. 3] The curation pipeline retains a text instance only when Qwen2.5-VL and OVIS2 produce identical transcriptions, and then discards all crops classified by a VLM as Level 1 or Level 2 (blurry or slightly blurry). The paper states this explicitly filters out 'misreadings, hard-to-read texts' and out-of-focus content. Since the SA-Text test set (Sec. 5.1) and Real-Text annotations are created with the same pipeline, the benchmark distribution systematically excludes the ambiguous, heavily degraded, or genuinely corrupted text that text restoration is most needed for. Consequently, the central claim that TeReDiff 'achieves the best F1-score at every level' (Sec. 5.2) and 'consistently outperforms state-of-the-art restoration methods' is established only for text that is legible enough for two strong VLMs to agree on before restoration. To support the real-world claim, the authors should either (a) evaluate on an independently annotated test set (e.g., manual transcriptions of RealSR/DRealSR crops that include hard cases), (b) report the fraction of instances removed at each filtering stage and compare method performance on retained vs. discarded instances, or (c) otherwise justify that the filter does not remove the cases where restoration is most needed.
  2. [Sec. 4.3, Tab. 5(a), Tab. 5(b)] The inference-time loop uses the text spotter's own output on the noisy image as the prompt for the next denoising step. The ablations show a large gap between ground-truth prompts and predicted prompts: for Level 2, the Stage3 F1-score is 71.85 with gt prompts versus 65.75 with predicted prompts (Tab. 5a), and the gap grows for Levels 1 and 3 in Tab. 9. This indicates that the benefit of the loop is bounded by the spotter's accuracy and that wrong recognized text can be fed back into the next step. The paper does not analyze error propagation over the denoising trajectory, nor does it quantify how often the final output text differs from the ground truth when the prompt is incorrect. Given that this feedback loop is a core component of the method, the authors should provide an analysis of prompt accuracy (e.g., word error rate of the prompter at early vs. late timesteps) and a qualitative discussion of cases where the prompt error compounds.
  3. [Sec. 5.2, Tab. 4] Image-quality metrics are reported only for DiffBIR (and its v2.1 variant), not for the other seven baselines (Real-ESRGAN, SwinIR, ResShift, StableSR, SeeSR, SUPIR, FaithDiff) that are compared in the text-spotting tables. Moreover, on SA-Text, TeReDiff is worse than DiffBIR on all four no-reference metrics (NIQE 5.452 vs. 5.107, MANIQA 0.6471 vs. 0.6771, MUSIQ 72.07 vs. 73.33, CLIPIQA 0.6145 vs. 0.6589), yet the text in Sec. 5.2 states the model performs 'comparably' on these metrics. This under-specification weakens the claim that TeReDiff restores 'without compromising overall image restoration performance' (Sec. 4.1, Sec. 5.2). The authors should report the same image-quality metrics for all baselines, or explicitly scope the claim to text-focused metrics and discuss the trade-off.
minor comments (5)
  1. [Sec. 4.4, Sec. C] Equation (2) defines λ_box for the box regression loss, Eq. (3) uses λ_poly for the polygon loss, and the implementation details then list λ_coord = 5.0 but no λ_poly or λ_box. Please unify this notation and list all hyperparameter values.
  2. [Appendix B] The claim that this is 'the first work to demonstrate that diffusion features are suitable for effectively learning text spotting' is likely too strong, given prior uses of diffusion features for OCR-related tasks (e.g., [33, 87] in the paper's own reference list). Please soften the claim and add comparative discussion.
  3. [Sec. 5.1, Tab. 2, Tab. 3] All text-spotting metrics are reported without error bars or multiple runs. Since the differences between TeReDiff and the second-best baseline are often small (e.g., Tab. 2, Level 1 TESTR F1: 67.47 vs. 66.23), statistical significance should be assessed or the results should be clearly labeled as single-run.
  4. [Sec. C] The implementation details state that 'two sets of queries ... are processed through cross-attention with the encoder output in the respectful decoder layers'—'respectful' should be 'respective'. Also, the prompter template concatenates all recognized texts into one prompt; the maximum number of text instances (K) and prompt truncation behavior should be described, since K can be large and the prompt model has a fixed context length (n=77 tokens).
  5. [Appendix D.3] The user study uses only 10 samples and 21 participants and compares only against DiffBIR; this is not a substitute for a full perceptual study, and the claims about human preference should be tempered accordingly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claims are empirical evaluations against fixed off-the-shelf text spotters, and the ablations separate predicted from ground-truth prompts, so the results are not definitionally forced.

full rationale

The paper's central claims are empirical: TeReDiff is trained on SA-Text and evaluated on held-out SA-Text and Real-Text using fixed, off-the-shelf text spotters (ABCNet v2 and TESTR). The recognition and detection metrics are not computed from the model's own prompt or text-spotting outputs; they come from external evaluators against VLM-consensus ground truth. The ablation study (Tab. 5) explicitly separates null, predicted, and ground-truth prompts and shows a large gap between predicted and ground-truth prompting (e.g., Stage3_pr F1 65.75 vs Stage3_gt 71.85), demonstrating that the reported result is not trivially equal to the training signal. The VLM-consensus curation pipeline is shared between training and test construction, and it does filter out ambiguous or blurry text; this is a legitimate concern about external validity and benchmark difficulty, but it is not a circular derivation: every baseline is evaluated on the same filtered benchmark and the comparison among methods remains well-defined. Self-citations in the related work (e.g., [11]-[13], [26]-[30]) are not load-bearing for the main result; no uniqueness theorem or prior claim by the same authors is invoked to forbid alternatives. No equation or metric reduces to a fitted parameter renamed as a prediction. Accordingly, the paper's derivation chain is self-contained with respect to its empirical claims, and the circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced; 'text-image hallucination' is a descriptive term for a known failure mode, not a new entity. The ledger instead tracks the hand-set hyperparameters, the VLM-based curation rule, and the domain assumptions about degradation realism and annotation accuracy that the central claim depends on.

free parameters (6)
  • Stage-3 loss weight lambda = 0.01
    Balances diffusion loss and text-spotting losses in Eq. 4; chosen by hand, not searched. It controls how strongly text supervision shapes restoration.
  • Text-spotting loss weights lambda_cls, lambda_coord, lambda_char, lambda_gIoU = 2.0, 5.0, 4.0, 2.0
    Hand-set weights in Eqs. 2 and 3; they affect detection and recognition supervision strength.
  • Confidence threshold T = 0.5
    Instances below T are dropped from text-spotting outputs and therefore from the prompt; this affects prompt quality and the number K of supported instances.
  • VLM agreement and sharpness filter = Identical transcriptions from two VLMs; keep only Level 3 sharpness
    Curation rule in Fig. 2 determines which text instances enter SA-Text and sets the ceiling for both training and test annotations.
  • Prompt template = 'A realistic scene where the texts ... appear clearly on signs, boards, buildings, or other objects.'
    Handcrafted caption style used at inference; the ablation in Tab. 5(b) shows prompt style affects predicted-prompt performance.
  • Number of inference sampling steps = 50
    Fixed inference setting; affects generation quality and runtime, chosen to match standard diffusion restoration practice.
assumptions (5)
  • domain assumption Real-ESRGAN degradation pipeline produces LQ images representative of real degradations.
    Used to synthesize all training and most test pairs (Sec. 5.1); if synthetic degradations diverge from field conditions, text restoration gains may not transfer.
  • domain assumption Diffusion features from SD2.1 U-Net carry enough text-aware information to train a text spotter.
    Central to the joint-training design (Sec. 4.2, App. B); validated only on SA-Text, not on other text benchmarks.
  • domain assumption Two VLM transcriptions that agree are correct enough to serve as ground truth.
    Dataset curation retains an instance only when Qwen2.5-VL and OVIS2 outputs match exactly (Sec. 3.1); agreement is treated as accuracy.
  • domain assumption Off-the-shelf ABCNet v2 and TESTR models reliably score restored text fidelity.
    These are the main evaluation metrics (Sec. 5.1); if recognizers are insensitive to specific character errors, reported recognition gains may overestimate fidelity.
  • domain assumption SA-1B crops provide high-quality HQ references.
    SA-Text is cropped from SA-1B and only VLM-sharpness filtered; there is no manual quality audit (Sec. 3.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Text-Aware Image Restoration with Diffusion Models." pith.science (2026). https://pith.science/paper/BDGTOJEN

@misc{pith2026250609993,
  author       = {Pith},
  title        = {Pith review of: Text-Aware Image Restoration with Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BDGTOJEN}},
  note         = {Machine review of arXiv:2506.09993}
}
read the original abstract

Image restoration aims to recover degraded images. However, existing diffusion-based restoration methods, despite great success in natural image restoration, often struggle to faithfully reconstruct textual regions in degraded images. Those methods frequently generate plausible but incorrect text-like patterns, a phenomenon we refer to as text-image hallucination. In this paper, we introduce Text-Aware Image Restoration (TAIR), a novel restoration task that requires the simultaneous recovery of visual contents and textual fidelity. To tackle this task, we present SA-Text, a large-scale benchmark of 100K high-quality scene images densely annotated with diverse and complex text instances. Furthermore, we propose a multi-task diffusion framework, called TeReDiff, that integrates internal features from diffusion models into a text-spotting module, enabling both components to benefit from joint training. This allows for the extraction of rich text representations, which are utilized as prompts in subsequent denoising steps. Extensive experiments demonstrate that our approach consistently outperforms state-of-the-art restoration methods, achieving significant gains in text recognition accuracy. See our project page: https://cvlab-kaist.github.io/TAIR/

Figures

Figures reproduced from arXiv: 2506.09993 by the authors.

Figure 1
Figure 1. Text-Aware Image Restoration (TAIR). Given a low-quality (LQ) image containing degraded text, our method faithfully restores the original textual content with high legibility and fidelity, whereas previous diffusion-based models [5, 94, 51, 9] often fail to recover the text regions. Abstract Image restoration aims to recover degraded images. However, existing diffusion￾based restoration methods, despite great succes… view at source ↗
Figure 2
Figure 2. SA-Text curation pipeline. First, a text-spotting model such as DG-Bridge Spotter [31] is applied to the entire image to detect text regions. Since detection at the full-image scale may fail to capture certain text instances, we further extract image patches corresponding to the detected regions and reapply the same model to each patch in order to detect potential false negatives. Next, two vision-language models (V… view at source ↗
Figure 3
Figure 3. Illustration of our dataset curation pipeline’s effectiveness. (a) Original high-resolution image with multiple text instances. (b) Ambiguous text instances are removed during the Vi￾sion–Language Model (VLM) filtering stage when the two VLMs produce differing recognition outputs. (c) Incorrect detections from the full image are corrected by re-running the detection model on smaller crops; here, the phrase "Powered … view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Example images from our SA-Text. Our dataset comprises high-quality, diverse im￾ages featuring text in varied sizes, styles, and layouts—including curved, rotated, and complex forms—providing a robust foundation for the proposed TAIR task. appears in only one crop. As …
Figure 5
Figure 5. Figure 5: Overview of the TeReDiff architecture, training, and inference pipeline. TAIR integrates a text-spotting module into a diffusion image restoration framework, using text supervision during training and recognized text as a prompt at inference to enhance text-aware image…
Figure 6
Figure 6. Figure 6: Example of textual prompt guidance for each timestep. The outputs of the text-spotting module serve as input prompts for the next timestep. The detected boxes and recognized texts may also change as the image is gradually restored over timesteps. 7 [PITH_FULL_IMAGE:fi…
Figure 7
Figure 7. Figure 7: Examples of images classified by blurriness. Images are categorized into Levels 1–3: Level 1 (very blurry), Level 2 (slightly blurry), and Level 3 (clearly focused). Images classified as Level 1 and Level 2 are excluded from the final dataset to ensure that only clearl…
Figure 8
Figure 8. Figure 8: Example samples for user study. E Additional Qualitative Results In [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Qualitative results on SA-Text test set Level 1. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: Qualitative results on SA-Text test set Level 2. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: Qualitative results on SA-Text test set Level 3. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_11.png]
Figure 12
Figure 12. Figure 12: Qualitative results on Real-Text. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

105 extracted references · 45 canonical work pages

  1. [1]

    Ntire 2017 challenge on single image super-resolution: Dataset and study

    Eirikur Agustsson and Radu Timofte. Ntire 2017 challenge on single image super-resolution: Dataset and study. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 126–135, 2017

    2017

  2. [2]

    Dreamclear: High-capacity real-world image restoration with privacy-safe dataset curation

    Yuang Ai, Xiaoqiang Zhou, Huaibo Huang, Xiaotian Han, Zhengyu Chen, Quanzeng You, and Hongxia Yang. Dreamclear: High-capacity real-world image restoration with privacy-safe dataset curation. Advances in Neural Information Processing Systems, 37:55443–55469, 2024

    2024

  3. [3]

    What is wrong with scene text recognition model compar- isons? dataset and model analysis

    Jeonghun Baek, Geewook Kim, Junyeop Lee, Sungrae Park, Dongyoon Han, Sangdoo Yun, Seong Joon Oh, and Hwalsuk Lee. What is wrong with scene text recognition model compar- isons? dataset and model analysis. In Proceedings of the IEEE/CVF international conference on computer vision, pages 4715–4723, 2019

    2019

  4. [4]

    Qwen2.5-vl technical report

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, Jiabo Ye, Xi Zhang, Tianbao Xie, Zesen Cheng, Hang Zhang, Zhibo Yang, Haiyang Xu, and Junyang Lin. Qwen2.5-vl technical report. a...

    arXiv 2025

  5. [5]

    Stable video diffusion: Scaling latent video diffusion models to large datasets

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Do- minik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127, 2023

    arXiv 2023

  6. [6]

    Toward real-world single image super-resolution: A new benchmark and a new model

    Jianrui Cai, Hui Zeng, Hongwei Yong, Zisheng Cao, and Lei Zhang. Toward real-world single image super-resolution: A new benchmark and a new model. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3086–3095, 2019

    2019

  7. [7]

    End-to-end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-end object detection with transformers. In European conference on computer vision, pages 213–229. Springer, 2020

    2020

  8. [8]

    Mode regularized generative adversarial networks

    Tong Che, Yanran Li, Athul Paul Jacob, Yoshua Bengio, and Wenjie Li. Mode regularized generative adversarial networks. arXiv preprint arXiv:1612.02136, 2016

    arXiv 2016

Show all 105 references
  1. [9]

    Faithdiff: Unleashing diffusion priors for faithful image super-resolution

    Junyang Chen, Jinshan Pan, and Jiangxin Dong. Faithdiff: Unleashing diffusion priors for faithful image super-resolution. arXiv preprint arXiv:2411.18824, 2024

    2024 arXiv

  2. [10]

    Total-text: A comprehensive dataset for scene text detection and recognition

    Chee Kheng Ch’ng and Chee Seng Chan. Total-text: A comprehensive dataset for scene text detection and recognition. In 2017 14th IAPR international conference on document analysis and recognition (ICDAR), volume 1, pages 935–942. IEEE, 2017

    2017

  3. [11]

    Cats: Cost aggregation transformers for visual correspondence

    Seokju Cho, Sunghwan Hong, Sangryul Jeon, Yunsung Lee, Kwanghoon Sohn, and Seungry- ong Kim. Cats: Cost aggregation transformers for visual correspondence. Advances in Neural Information Processing Systems, 34:9011–9023, 2021

    2021

  4. [12]

    Cats++: Boosting cost aggregation with convolutions and transformers

    Seokju Cho, Sunghwan Hong, and Seungryong Kim. Cats++: Boosting cost aggregation with convolutions and transformers. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(6):7174–7194, 2022

    2022

  5. [13]

    Cat-seg: Cost aggregation for open-vocabulary semantic segmentation

    Seokju Cho, Heeseong Shin, Sunghwan Hong, Anurag Arnab, Paul Hongsuck Seo, and Seungryong Kim. Cat-seg: Cost aggregation for open-vocabulary semantic segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4113–4123, 2024

    2024

  6. [14]

    Image quality assessment: Unifying structure and texture similarity

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simoncelli. Image quality assessment: Unifying structure and texture similarity. IEEE transactions on pattern analysis and machine intelligence, 44(5):2567–2581, 2020

    2020

  7. [15]

    Boosting optical character recognition: A super-resolution approach

    Chao Dong, Ximei Zhu, Yubin Deng, Chen Change Loy, and Yu Qiao. Boosting optical character recognition: A super-resolution approach. arXiv preprint arXiv:1506.02211, 2015. 22

    2015 arXiv

  8. [16]

    Svtr: Scene text recognition with a single visual model

    Yongkun Du, Zhineng Chen, Caiyan Jia, Xiaoting Yin, Tianlun Zheng, Chenxia Li, Yuning Du, and Yu-Gang Jiang. Svtr: Scene text recognition with a single visual model. arXiv preprint arXiv:2205.00159, 2022

    2022 arXiv

  9. [17]

    The mapillary traffic sign dataset for detection and classification on a global scale

    Christian Ertler, Jerneja Mislej, Tobias Ollmann, Lorenzo Porzi, Gerhard Neuhold, and Yubin Kuang. The mapillary traffic sign dataset for detection and classification on a global scale. In European Conference on Computer Vision, pages 68–84. Springer, 2020

    2020

  10. [18]

    Read like humans: Autonomous, bidirectional and iterative language modeling for scene text recognition

    Shancheng Fang, Hongtao Xie, Yuxin Wang, Zhendong Mao, and Yongdong Zhang. Read like humans: Autonomous, bidirectional and iterative language modeling for scene text recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7098–7...

    2021

  11. [19]

    Deep unrestricted document image rectification

    Hao Feng, Shaokai Liu, Jiajun Deng, Wengang Zhou, and Houqiang Li. Deep unrestricted document image rectification. IEEE Transactions on Multimedia, 26:6142–6154, 2023

    2023

  12. [20]

    Docscanner: Robust document image rectification with progressive learning

    Hao Feng, Wengang Zhou, Jiajun Deng, Qi Tian, and Houqiang Li. Docscanner: Robust document image rectification with progressive learning. International Journal of Computer Vision, pages 1–20, 2025

    2025

  13. [21]

    Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks

    Alex Graves, Santiago Fernández, Faustino Gomez, and Jürgen Schmidhuber. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks. In Proceedings of the 23rd international conference on Machine learning , pages 369–376, 2006

    2006

  14. [22]

    Synthetic data for text localisation in natural images

    Ankush Gupta, Andrea Vedaldi, and Andrew Zisserman. Synthetic data for text localisation in natural images. In IEEE Conference on Computer Vision and Pattern Recognition, 2016

    2016

  15. [23]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  16. [24]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems, 30, 2017

    2017

  17. [25]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Annual Conference on Neural Information Processing Systems, 2020

    2020

  18. [26]

    Cost aggregation with 4d convolutional swin transformer for few-shot segmentation

    Sunghwan Hong, Seokju Cho, Jisu Nam, Stephen Lin, and Seungryong Kim. Cost aggregation with 4d convolutional swin transformer for few-shot segmentation. In European Conference on Computer Vision, pages 108–126. Springer, 2022

    2022

  19. [27]

    Pf3plat: Pose-free feed-forward 3d gaussian splatting

    Sunghwan Hong, Jaewoo Jung, Heeseong Shin, Jisang Han, Jiaolong Yang, Chong Luo, and Seungryong Kim. Pf3plat: Pose-free feed-forward 3d gaussian splatting. arXiv preprint arXiv:2410.22128, 2024

    2024 arXiv

  20. [28]

    Unifying correspondence pose and nerf for generalized pose-free novel view synthesis

    Sunghwan Hong, Jaewoo Jung, Heeseong Shin, Jiaolong Yang, Seungryong Kim, and Chong Luo. Unifying correspondence pose and nerf for generalized pose-free novel view synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20196–20206, 2024

    2024

  21. [29]

    Deep matching prior: Test-time optimization for dense correspondence

    Sunghwan Hong and Seungryong Kim. Deep matching prior: Test-time optimization for dense correspondence. In Proceedings of the IEEE/CVF international conference on computer vision, pages 9907–9917, 2021

    2021

  22. [30]

    Neural matching fields: Implicit representation of matching fields for visual correspondence

    Sunghwan Hong, Jisu Nam, Seokju Cho, Susung Hong, Sangryul Jeon, Dongbo Min, and Seungryong Kim. Neural matching fields: Implicit representation of matching fields for visual correspondence. Advances in Neural Information Processing Systems, 35:13512–13526, 2022

    2022

  23. [31]

    Bridging the gap between end-to-end and two-step text spotting

    Mingxin Huang, Hongliang Li, Yuliang Liu, Xiang Bai, and Lianwen Jin. Bridging the gap between end-to-end and two-step text spotting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15608–15618, 2024. 23

    2024

  24. [32]

    Icdar 2015 competition on robust reading

    Dimosthenis Karatzas, Lluis Gomez-Bigorda, Anguelos Nicolaou, Suman Ghosh, Andrew Bagdanov, Masakazu Iwamura, Jiri Matas, Lukas Neumann, Vijay Ramaseshan Chandrasekhar, Shijian Lu, et al. Icdar 2015 competition on robust reading. In 2015 13th international conference on docume...

    2015

  25. [33]

    Repurposing diffusion-based image generators for monocular depth estimation

    Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Metzger, Rodrigo Caye Daudt, and Kon- rad Schindler. Repurposing diffusion-based image generators for monocular depth estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9492–9502, 2024

    2024

  26. [34]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pages 5148–5157, 2021

    2021

  27. [35]

    Auto-encoding variational bayes, 2013

    Diederik P Kingma, Max Welling, et al. Auto-encoding variational bayes, 2013

    2013

  28. [36]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. In Proceedings of the IEEE/CVF international conference on computer vision, pages 4015–4026, 2023

    2023

  29. [37]

    To- wards weakly-supervised text spotting using a multi-task transformer

    Yair Kittenplon, Inbal Lavi, Sharon Fogel, Yarin Bar, R Manmatha, and Pietro Perona. To- wards weakly-supervised text spotting using a multi-task transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4604–4613, 2022

    2022

  30. [38]

    The hungarian method for the assignment problem

    Harold W Kuhn. The hungarian method for the assignment problem. Naval research logistics quarterly, 2(1-2):83–97, 1955

    1955

  31. [39]

    Photo-realistic single image super-resolution using a generative adversarial network

    Christian Ledig, Lucas Theis, Ferenc Huszár, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, et al. Photo-realistic single image super-resolution using a generative adversarial network. In Proceedings of the IEEE c...

    2017

  32. [40]

    On recognizing texts of arbitrary shapes with 2d self-attention

    Junyeop Lee, Sungrae Park, Jeonghun Baek, Seong Joon Oh, Seonghyeon Kim, and Hwalsuk Lee. On recognizing texts of arbitrary shapes with 2d self-attention. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 546–547, 2020

    2020

  33. [41]

    Textslam: Visual slam with planar text features

    Boying Li, Danping Zou, Daniele Sartori, Ling Pei, and Wenxian Yu. Textslam: Visual slam with planar text features. In 2020 IEEE International Conference on Robotics and Automation (ICRA), pages 2102–2108. IEEE, 2020

    2020

  34. [42]

    Foreground and text-lines aware document image rectification

    Heng Li, Xiangping Wu, Qingcai Chen, and Qianjin Xiang. Foreground and text-lines aware document image rectification. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 19574–19583, 2023

    2023

  35. [43]

    Show, attend and read: A simple and strong baseline for irregular text recognition

    Hui Li, Peng Wang, Chunhua Shen, and Guyu Zhang. Show, attend and read: A simple and strong baseline for irregular text recognition. In Proceedings of the AAAI conference on artificial intelligence, volume 33, pages 8610–8617, 2019

    2019

  36. [44]

    Lsdir: A large scale dataset for image restoration

    Yawei Li, Kai Zhang, Jingyun Liang, Jiezhang Cao, Ce Liu, Rui Gong, Yulun Zhang, Hao Tang, Yun Liu, Denis Demandolx, et al. Lsdir: A large scale dataset for image restoration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1775–1787, 2023

    2023

  37. [45]

    Swinir: Image restoration using swin transformer

    Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. Swinir: Image restoration using swin transformer. In Proceedings of the IEEE/CVF interna- tional conference on computer vision, pages 1833–1844, 2021

    2021

  38. [46]

    Dasr: Dual-attention transformer for infrared image super-resolution

    ShuBo Liang, Kechen Song, Wenli Zhao, Song Li, and Yunhui Yan. Dasr: Dual-attention transformer for infrared image super-resolution. Infrared Physics & Technology, 133:104837, 2023. 24

    2023

  39. [47]

    Textboxes: A fast text detector with a single deep neural network

    Minghui Liao, Baoguang Shi, Xiang Bai, Xinggang Wang, and Wenyu Liu. Textboxes: A fast text detector with a single deep neural network. In Proceedings of the AAAI conference on artificial intelligence, volume 31, 2017

    2017

  40. [48]

    Real-time scene text detection with differentiable binarization

    Minghui Liao, Zhaoyi Wan, Cong Yao, Kai Chen, and Xiang Bai. Real-time scene text detection with differentiable binarization. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 11474–11481, 2020

    2020

  41. [49]

    Enhanced deep residual networks for single image super-resolution

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. Enhanced deep residual networks for single image super-resolution. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops, pages 136–144, 2017

    2017

  42. [50]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. Focal loss for dense object detection. In Proceedings of the IEEE international conference on computer vision, pages 2980–2988, 2017

    2017

  43. [51]

    Diffbir: Toward blind image restoration with generative diffusion prior

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Bo Dai, Fanghua Yu, Yu Qiao, Wanli Ouyang, and Chao Dong. Diffbir: Toward blind image restoration with generative diffusion prior. In European Conference on Computer Vision, pages 430–448. Springer, 2024

    2024

  44. [52]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. Advances in neural information processing systems, 36:34892–34916, 2023

    2023

  45. [53]

    Abcnet: Real-time scene text spotting with adaptive bezier-curve network

    Yuliang Liu, Hao Chen, Chunhua Shen, Tong He, Lianwen Jin, and Liangwei Wang. Abcnet: Real-time scene text spotting with adaptive bezier-curve network. In proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9809–9818, 2020

    2020

  46. [54]

    Deep matching prior network: Toward tighter multi-oriented text detection

    Yuliang Liu and Lianwen Jin. Deep matching prior network: Toward tighter multi-oriented text detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1962–1969, 2017

    1962

  47. [55]

    On the hidden mystery of ocr in large multimodal models

    Yuliang Liu, Zhang Li, Biao Yang, Chunyuan Li, Xucheng Yin, Cheng-lin Liu, Lianwen Jin, and Xiang Bai. On the hidden mystery of ocr in large multimodal models. arXiv e-prints, pages arXiv–2305, 2023

    2023

  48. [56]

    Abcnet v2: Adaptive bezier-curve network for real-time end-to-end text spotting

    Yuliang Liu, Chunhua Shen, Lianwen Jin, Tong He, Peng Chen, Chongyu Liu, and Hao Chen. Abcnet v2: Adaptive bezier-curve network for real-time end-to-end text spotting. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(11):8048–8064, 2021

    2021

  49. [57]

    Ovis: Structural embedding alignment for multimodal large language model

    Shiyin Lu, Yang Li, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang, and Han- Jia Ye. Ovis: Structural embedding alignment for multimodal large language model. arXiv:2405.20797, 2024

    2024 arXiv

  50. [58]

    Arbitrary-oriented scene text detection via rotation proposals

    Jianqi Ma, Weiyuan Shao, Hao Ye, Li Wang, Hong Wang, Yingbin Zheng, and Xiangyang Xue. Arbitrary-oriented scene text detection via rotation proposals. IEEE transactions on multimedia, 20(11):3111–3122, 2018

    2018

  51. [59]

    Docunet: Document image unwarping via a stacked u-net

    Ke Ma, Zhixin Shu, Xue Bai, Jue Wang, and Dimitris Samaras. Docunet: Document image unwarping via a stacked u-net. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4700–4709, 2018

    2018

  52. [60]

    Mode seeking generative adversarial networks for diverse image synthesis

    Qi Mao, Hsin-Ying Lee, Hung-Yu Tseng, Siwei Ma, and Ming-Hsuan Yang. Mode seeking generative adversarial networks for diverse image synthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1429–1437, 2019

    2019

  53. [61]

    Codi: conditional diffusion distillation for higher-fidelity and faster image generation

    Kangfu Mei, Mauricio Delbracio, Hossein Talebi, Zhengzhong Tu, Vishal M Patel, and Peyman Milanfar. Codi: conditional diffusion distillation for higher-fidelity and faster image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, p...

    2024

  54. [62]

    Sdxl: Improving latent diffusion models for high-resolution image synthesis

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023. 25

    2023 arXiv

  55. [63]

    Dntextspotter: Arbitrary-shaped scene text spotting via improved denoising training

    Qian Qiao, Yu Xie, Jun Gao, Tianxiang Wu, Shaoyao Huang, Jiaqing Fan, Ziqiang Cao, Zili Wang, and Yue Zhang. Dntextspotter: Arbitrary-shaped scene text spotting via improved denoising training. In Proceedings of the 32nd ACM International Conference on Multimedia, pages 10134–...

    2024

  56. [64]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International conference on machine learning, pages 8821–8831. Pmlr, 2021

    2021

  57. [65]

    Learning to localize objects improves spatial reasoning in visual-llms

    Kanchana Ranasinghe, Satya Narayan Shukla, Omid Poursaeed, Michael S Ryoo, and Tsung- Yu Lin. Learning to localize objects improves spatial reasoning in visual-llms. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 12977– 12987, 2024

    2024

  58. [66]

    Generalized intersection over union: A metric and a loss for bounding box regression

    Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian Reid, and Silvio Savarese. Generalized intersection over union: A metric and a loss for bounding box regression. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 658–666, 2019

    2019

  59. [67]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022

    2022

  60. [68]

    Denoising diffusion proba- bilistic models for robust image super-resolution in the wild

    Hshmat Sahak, Daniel Watson, Chitwan Saharia, and David Fleet. Denoising diffusion proba- bilistic models for robust image super-resolution in the wild. arXiv preprint arXiv:2302.07864, 2023

    2023 arXiv

  61. [69]

    Image super-resolution via iterative refinement

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J Fleet, and Mohammad Norouzi. Image super-resolution via iterative refinement. IEEE transactions on pattern analysis and machine intelligence, 45(4):4713–4726, 2022

    2022

  62. [70]

    An end-to-end trainable neural network for image- based sequence recognition and its application to scene text recognition

    Baoguang Shi, Xiang Bai, and Cong Yao. An end-to-end trainable neural network for image- based sequence recognition and its application to scene text recognition. IEEE transactions on pattern analysis and machine intelligence, 39(11):2298–2304, 2016

    2016

  63. [71]

    Aster: An attentional scene text recognizer with flexible rectification

    Baoguang Shi, Mingkun Yang, Xinggang Wang, Pengyuan Lyu, Cong Yao, and Xiang Bai. Aster: An attentional scene text recognizer with flexible rectification. IEEE transactions on pattern analysis and machine intelligence, 41(9):2035–2048, 2018

    2018

  64. [72]

    Textocr: Towards large-scale end-to-end reasoning for arbitrary-shaped scene text

    Amanpreet Singh, Guan Pang, Mandy Toh, Jing Huang, Wojciech Galuba, and Tal Hassner. Textocr: Towards large-scale end-to-end reasoning for arbitrary-shaped scene text. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 8802–8812, 2021

    2021

  65. [73]

    Deep learning for large-scale traffic-sign detection and recognition

    Domen Tabernik and Danijel Skoˇcaj. Deep learning for large-scale traffic-sign detection and recognition. IEEE transactions on intelligent transportation systems, 21(4):1427–1440, 2019

    2019

  66. [74]

    Few could be better than all: Feature sampling and grouping for scene text detection

    Jingqun Tang, Wenqing Zhang, Hongye Liu, MingKun Yang, Bo Jiang, Guanglong Hu, and Xiang Bai. Few could be better than all: Feature sampling and grouping for scene text detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4563–...

    2022

  67. [75]

    Emergent correspondence from image diffusion

    Luming Tang, Menglin Jia, Qianqian Wang, Cheng Perng Phoo, and Bharath Hariharan. Emergent correspondence from image diffusion. In Thirty-seventh Conference on Neural Information Processing Systems, 2023

    2023

  68. [76]

    Coco-text: Dataset and benchmark for text detection and recognition in natural images

    Andreas Veit, Tomas Matera, Lukas Neumann, Jiri Matas, and Serge Belongie. Coco-text: Dataset and benchmark for text detection and recognition in natural images. arXiv preprint arXiv:1601.07140, 2016

    2016 arXiv

  69. [77]

    Exploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Exploring clip for assessing the look and feel of images. In Proceedings of the AAAI conference on artificial intelligence, volume 37, pages 2555–2563, 2023. 26

    2023

  70. [78]

    Exploiting diffusion prior for real-world image super-resolution

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin CK Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution. International Journal of Computer Vision, 132(12):5929–5949, 2024

    2024

  71. [79]

    Shape ro- bust text detection with progressive scale expansion network

    Wenhai Wang, Enze Xie, Xiang Li, Wenbo Hou, Tong Lu, Gang Yu, and Shuai Shao. Shape ro- bust text detection with progressive scale expansion network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9336–9345, 2019

    2019

  72. [80]

    Scene text image super-resolution in the wild

    Wenjia Wang, Enze Xie, Xuebo Liu, Wenhai Wang, Ding Liang, Chunhua Shen, and Xiang Bai. Scene text image super-resolution in the wild. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part X 16 , pages 650–666. Springer, 2020

    2020

  73. [81]

    Textsr: Content-aware text super-resolution guided by recognition

    Wenjia Wang, Enze Xie, Peize Sun, Wenhai Wang, Lixun Tian, Chunhua Shen, and Ping Luo. Textsr: Content-aware text super-resolution guided by recognition. arXiv preprint arXiv:1909.07113, 2019

    1909 arXiv

  74. [82]

    Real-esrgan: Training real-world blind super-resolution with pure synthetic data

    Xintao Wang, Liangbin Xie, Chao Dong, and Ying Shan. Real-esrgan: Training real-world blind super-resolution with pure synthetic data. In Proceedings of the IEEE/CVF international conference on computer vision, pages 1905–1914, 2021

    1905

  75. [83]

    Esrgan: Enhanced super-resolution generative adversarial networks

    Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy. Esrgan: Enhanced super-resolution generative adversarial networks. In Pro- ceedings of the European conference on computer vision (ECCV) workshops , pages 0–0, 2018

    2018

  76. [84]

    Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600– 612, 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600– 612, 2004

    2004

  77. [85]

    Component divide-and-conquer for real-world image super-resolution

    Pengxu Wei, Ziwei Xie, Hannan Lu, Zongyuan Zhan, Qixiang Ye, Wangmeng Zuo, and Liang Lin. Component divide-and-conquer for real-world image super-resolution. In Com- puter Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part VIII 16, p...

    2020

  78. [86]

    Seesr: Towards semantics-aware real-world image super-resolution

    Rongyuan Wu, Tao Yang, Lingchen Sun, Zhengqiang Zhang, Shuai Li, and Lei Zhang. Seesr: Towards semantics-aware real-world image super-resolution. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 25456–25467, 2024

    2024

  79. [87]

    Open-vocabulary panoptic segmentation with text-to-image diffusion models

    Jiarui Xu, Sifei Liu, Arash Vahdat, Wonmin Byeon, Xiaolong Wang, and Shalini De Mello. Open-vocabulary panoptic segmentation with text-to-image diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2955–2966, 2023

    2023

  80. [88]

    Learning to super-resolve blurry face and text images

    Xiangyu Xu, Deqing Sun, Jinshan Pan, Yujin Zhang, Hanspeter Pfister, and Ming-Hsuan Yang. Learning to super-resolve blurry face and text images. In Proceedings of the IEEE international conference on computer vision, pages 251–260, 2017

    2017

  81. [89]

    Maniqa: Multi-dimension attention network for no-reference image quality assessment

    Sidi Yang, Tianhe Wu, Shuwei Shi, Shanshan Lao, Yuan Gong, Mingdeng Cao, Jiahao Wang, and Yujiu Yang. Maniqa: Multi-dimension attention network for no-reference image quality assessment. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, page...

    2022

  82. [90]

    Cogvideox: Text-to-video diffusion models with an expert transformer

    Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, Yuanming Yang, Wenyi Hong, Xiaohan Zhang, Guanyu Feng, et al. Cogvideox: Text-to-video diffusion models with an expert transformer. arXiv preprint arXiv:2408.06072, 2024

    2024 arXiv

  83. [91]

    Dptext-detr: Towards better scene text detection with dynamic points in transformer

    Maoyuan Ye, Jing Zhang, Shanshan Zhao, Juhua Liu, Bo Du, and Dacheng Tao. Dptext-detr: Towards better scene text detection with dynamic points in transformer. In Proceedings of the AAAI conference on artificial intelligence, volume 37, pages 3241–3249, 2023. 27

    2023

  84. [92]

    Deepsolo: Let transformer decoder with explicit points solo for text spotting

    Maoyuan Ye, Jing Zhang, Shanshan Zhao, Juhua Liu, Tongliang Liu, Bo Du, and Dacheng Tao. Deepsolo: Let transformer decoder with explicit points solo for text spotting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19348–19357, 2023

    2023

  85. [93]

    Towards accurate scene text recognition with semantic reasoning networks

    Deli Yu, Xuan Li, Chengquan Zhang, Tao Liu, Junyu Han, Jingtuo Liu, and Errui Ding. Towards accurate scene text recognition with semantic reasoning networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12113–12122, 2020

    2020

  86. [94]

    Scaling up to excellence: Practicing model scaling for photo-realistic image restoration in the wild

    Fanghua Yu, Jinjin Gu, Zheyuan Li, Jinfan Hu, Xiangtao Kong, Xintao Wang, Jingwen He, Yu Qiao, and Chao Dong. Scaling up to excellence: Practicing model scaling for photo-realistic image restoration in the wild. In Proceedings of the IEEE/CVF Conference on Computer Vision and ...

    2024

  87. [95]

    Resshift: Efficient diffusion model for image super-resolution by residual shifting

    Zongsheng Yue, Jianyi Wang, and Chen Change Loy. Resshift: Efficient diffusion model for image super-resolution by residual shifting. Advances in Neural Information Processing Systems, 36:13294–13307, 2023

    2023

  88. [96]

    Detecting curve text in the wild: New dataset and new solution

    Liu Yuliang, Jin Lianwen, Zhang Shuaitao, and Zhang Sheng. Detecting curve text in the wild: New dataset and new solution. arXiv preprint arXiv:1712.02170, 2017

    2017 arXiv

  89. [97]

    Blind image super-resolution via contrastive representation learning

    Jiahui Zhang, Shijian Lu, Fangneng Zhan, and Yingchen Yu. Blind image super-resolution via contrastive representation learning. arXiv preprint arXiv:2107.00708, 2021

    2021 arXiv

  90. [98]

    A feature-enriched completely blind image quality evaluator

    Lin Zhang, Lei Zhang, and Alan C Bovik. A feature-enriched completely blind image quality evaluator. IEEE Transactions on Image Processing, 24(8):2579–2591, 2015

    2015

  91. [99]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023

    2023

  92. [100]

    The unrea- sonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unrea- sonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 586–595, 2018

    2018

  93. [101]

    Text spotting transformers

    Xiang Zhang, Yongwen Su, Subarna Tripathi, and Zhuowen Tu. Text spotting transformers. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9519–9528, 2022

    2022

  94. [102]

    Zoom to learn, learn to zoom

    Xuaner Zhang, Qifeng Chen, Ren Ng, and Vladlen Koltun. Zoom to learn, learn to zoom. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3762–3770, 2019

    2019

  95. [103]

    Open-sora: Democratizing efficient video production for all

    Zangwei Zheng, Xiangyu Peng, Tianji Yang, Chenhui Shen, Shenggui Li, Hongxin Liu, Yukun Zhou, Tianyi Li, and Yang You. Open-sora: Democratizing efficient video production for all. arXiv preprint arXiv:2412.20404, 2024

    2024 arXiv

  96. [104]

    Deformable detr: Deformable transformers for end-to-end object detection

    Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. Deformable detr: Deformable transformers for end-to-end object detection. arXiv preprint arXiv:2010.04159, 2020

    2010 arXiv

  97. [105]

    Traffic-sign detection and classification in the wild

    Zhe Zhu, Dun Liang, Songhai Zhang, Xiaolei Huang, Baoli Li, and Shimin Hu. Traffic-sign detection and classification in the wild. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2110–2118, 2016. 28

    2016

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.