UniCLIP: Unified Framework for Contrastive Language-Image Pre-training

Authors: Janghyeon Lee, Jongsuk Kim, Hyounguk Shon, Bumsoo Kim, Seung Hwan Kim, Honglak Lee, Junmo Kim

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In our experiments, we show that each component that comprises Uni CLIP contributes well to the final performance.
Researcher Affiliation Collaboration Janghyeon Lee LG AI Research janghyeon.lee@lgresearch.ai Jongsuk Kim KAIST jskpop@kaist.ac.kr
Pseudocode No The paper describes methods using text and mathematical formulas but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code No Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No]
Open Datasets Yes Datasets For reproducibility, we use publicly available datasets for training and evaluation in our experiments, including CC3M [23], CC12M [2], De CLIP YFCC15M [13, 25] for training and Pets [18], CIFAR-10, CIFAR-100 [12], SUN397 [29], Food-101 [1], Flowers [17], Cars [11], Caltech101 [8], Aircraft [14], DTD [6], Image Net-1k [21], Flickr30k [19], COCO Captions [5] for evaluation.
Dataset Splits Yes Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes]
Hardware Specification No The paper's main text does not specify specific hardware details such as GPU/CPU models or types of internal clusters/cloud providers used for experiments, although the ethics statement indicates this information was provided.
Software Dependencies No The paper does not provide specific version numbers for ancillary software dependencies (e.g., 'Python 3.8, PyTorch 1.9').
Experiment Setup Yes Settings For each original image and corresponding text caption, one weakly augmented image, two strongly augmented images, and one text form a positive group in our experiments. Detailed augmentation and optimization configurations can be found in Appendix.