Long-tailed Diffusion Models with Oriented Calibration
Authors: Tianjiao Zhang, Huangjie Zheng, Jiangchao Yao, Xiangfeng Wang, Mingyuan Zhou, Ya Zhang, Yanfeng Wang
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We extensively evaluate our approach with experiments on multiple benchmark datasets, demonstrating its effectiveness and superior performance compared to existing methods. |
| Researcher Affiliation | Collaboration | 1 Cooperative Medianet Innovation Center, Shanghai Jiao Tong University 2 Shanghai Artificial Intelligence Laboratory 3 The University of Texas at Austin 4 East China Normal University |
| Pseudocode | Yes | Algorithm 1 T2H algorithm for conditional long tail generation |
| Open Source Code | Yes | Code: https://github.com/Media Brain-SJTU/OC_LT. |
| Open Datasets | Yes | We started by selecting two widely utilized datasets in the field of image synthesis, namely CIFAR10/CIFAR100, with their long-tailed versions CIFAR10LT and CIFAR100LT. and We also conduct experiments on a dataset Tiny Image Net200LT with more classes and higher resolution, which is the long tail version of Tiny Image Net200 (Tavanaei, 2020). |
| Dataset Splits | Yes | During the inference time, we generate 50k images for the evaluation of the metrics. and For the conditional generation, there is label information injected into the diffusion model in the training stage. While at the inference stage, the diffusion model is asked to generate 50k/L images for each class where L is the number of classes. and The metric is based on 10k generated images referenced with its validation set, as shown in Table. 3 and We generate 20k images with 1000 classes and use the balanced validation set with 20k images as the reference set for the calculation of FID scores. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU models, CPU types) used for running the experiments. |
| Software Dependencies | No | The paper mentions using a 'DDIM sampler' and following 'DDPM settings' and 'CBDM' implementation, but does not specify software dependencies with version numbers (e.g., Python, PyTorch/TensorFlow versions). |
| Experiment Setup | Yes | Our training schedules are strictly follows the implementation of CBDM (Qin et al., 2023), which follows the DDPM settings. and A DDIM sampler (Song et al., 2020a) is utilized with 100 steps of 10 steps skip comparison with initial DDPM (Ho et al., 2020) 1000 steps. |