Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
AnimateQR: Bridging Aesthetics and Functionality in Dynamic QR Code Generation
Authors: Guangyang Wu, Huayu Zheng, Siqi Luo, Guangtao Zhai, Xiaohong Liu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results demonstrate that Animate QR achieves state-of-the-art performance in both decoding success rates (96% vs. 56% baseline) and visual quality (user preference: 7.2 vs. 2.3 on a 10-point scale). |
| Researcher Affiliation | Academia | Guangyang Wu, Huayu Zheng, Siqi Luo, Guangtao Zhai, Xiaohong Liu Shanghai Jiao Tong University EMAIL |
| Pseudocode | Yes | Algorithm 1 Progressive Spatiotemporal Control (Pro ST) |
| Open Source Code | Yes | Codes are availble at https://github.com/mulns/Animate QR. |
| Open Datasets | No | For training the HLG-Control Net, we adopt Stable Diffusion v1.5 as the backbone and utilize a dataset comprising 60,000 high-resolution images, each preprocessed to a resolution of 512 × 512 pixels. For comparative evaluation, we construct a dataset of 500 uniquely stylized QR images, each with a resolution of 1024 × 1024 pixels, encompassing diverse visual content and artistic styles. No specific access information (link, DOI, or formal citation) is provided for these datasets. |
| Dataset Splits | No | For training the HLG-Control Net, we adopt Stable Diffusion v1.5 as the backbone and utilize a dataset comprising 60,000 high-resolution images, each preprocessed to a resolution of 512 × 512 pixels. We generate a set of 500 uniquely stylized QR images... We perform side-by-side comparisons on 50 QR code sequences... The paper mentions using these datasets for training and evaluation but does not specify any explicit training/test/validation splits for the 60,000 images, nor for the 500 or 50 images used in comparative evaluations. |
| Hardware Specification | Yes | Our implementation is based on the PyTorch framework and runs on an NVIDIA GeForce RTX 4090 GPU. |
| Software Dependencies | No | Our implementation is based on the PyTorch framework and runs on an NVIDIA GeForce RTX 4090 GPU. We adopt Stable Diffusion v1.5 as the backbone and utilize a dataset comprising 60,000 high-resolution images... Additionally, we employ Control Net [43] for spatial luminance control and Animate Diff [9] for motion generation in video sequences. The paper mentions 'PyTorch framework' without a specific version number and 'Animate Diff' without a version for the main method (only variants in Table 7). While 'Stable Diffusion v1.5' has a version, the criteria require multiple key software components with versions or a self-contained solver with a version. |
| Experiment Setup | Yes | We set the Control Net control strength to 0.9, the number of frames in Animate Diff to 16, and the motion scale to 1.0. By default, we define the keyframe set as K = {1, 8, 16} and set the learning rate to η = 0.001. |