T-MARS: Improving Visual Representations by Circumventing Text Feature Learning

• URL: http://arxiv.org/abs/2307.03132v2
• Date: Mon, 18 Mar 2024 22:18:02 GMT
• Title: T-MARS: Improving Visual Representations by Circumventing Text Feature Learning
• Authors: Pratyush Maini, Sachin Goyal, Zachary C. Lipton, J. Zico Kolter, Aditi Raghunathan,
• Abstract summary: We propose a new data filtering approach motivated by our observation that nearly 40% of LAION's images contain text that overlaps significantly with the caption. Our simple and scalable approach, T-MARS, filters out only those pairs where the text dominates the remaining visual features.
• Score: 99.3682210827572
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large web-sourced multimodal datasets have powered a slew of new methods for learning general-purpose visual representations, advancing the state of the art in computer vision and revolutionizing zero- and few-shot recognition. One crucial decision facing practitioners is how, if at all, to curate these ever-larger datasets. For example, the creators of the LAION-5B dataset chose to retain only image-caption pairs whose CLIP similarity score exceeded a designated threshold. In this paper, we propose a new state-of-the-art data filtering approach motivated by our observation that nearly 40% of LAION's images contain text that overlaps significantly with the caption. Intuitively, such data could be wasteful as it incentivizes models to perform optical character recognition rather than learning visual features. However, naively removing all such data could also be wasteful, as it throws away images that contain visual features (in addition to overlapping text). Our simple and scalable approach, T-MARS (Text Masking and Re-Scoring), filters out only those pairs where the text dominates the remaining visual features -- by first masking out the text and then filtering out those with a low CLIP similarity score of the masked image. Experimentally, T-MARS outperforms the top-ranked method on the "medium scale" of DataComp (a data filtering benchmark) by a margin of 6.5% on ImageNet and 4.7% on VTAB. Additionally, our systematic evaluation on various data pool sizes from 2M to 64M shows that the accuracy gains enjoyed by T-MARS linearly increase as data and compute are scaled exponentially. Code is available at https://github.com/locuslab/T-MARS.

Related papers

• VEGA: Learning Interleaved Image-Text Comprehension in Vision-Language Large Models [76.94378391979228]
  We introduce a new, more demanding task known as Interleaved Image-Text (IITC) This task challenges models to discern and disregard superfluous elements in both images and text to accurately answer questions. In support of this task, we further craft a new VEGA dataset, tailored for the IITC task on scientific content, and devised a subtask, Image-Text Association (ITA)
  arXiv  Detail & Related papers  (2024-06-14T17:59:40Z)
• Text Data-Centric Image Captioning with Interactive Prompts [20.48013600818985]
  Supervised image captioning approaches have made great progress, but it is challenging to collect high-quality human-annotated image-text data. This paper proposes a new Text data-centric approach with Interactive Prompts for image Captioning, named TIPCap.
  arXiv  Detail & Related papers  (2024-03-28T07:43:49Z)
• Learning Vision from Models Rivals Learning Vision from Data [54.43596959598465]
  We introduce SynCLR, a novel approach for learning visual representations exclusively from synthetic images and synthetic captions. We synthesize a large dataset of image captions using LLMs, then use an off-the-shelf text-to-image model to generate multiple images corresponding to each synthetic caption. We perform visual representation learning on these synthetic images via contrastive learning, treating images sharing the same caption as positive pairs.
  arXiv  Detail & Related papers  (2023-12-28T18:59:55Z)
• Sieve: Multimodal Dataset Pruning Using Image Captioning Models [11.362835828985494]
  Vision-Language Models (VLMs) are pretrained on large, diverse, and noisy web-crawled datasets. We argue that this approach suffers from multiple limitations including false positives and negatives due to CLIP's pretraining on noisy labels. We propose a pruning signal, Sieve, that employs synthetic captions generated by image-captioning models pretrained on small, diverse, and well-aligned image-text pairs.
  arXiv  Detail & Related papers  (2023-10-03T14:53:53Z)
• Improving Multimodal Datasets with Image Captioning [65.74736570293622]
  We study how generated captions can increase the utility of web-scraped datapoints with nondescript text. Our experiments with using generated captions at DataComp's large scale (1.28B image-text pairs) offer insights into the limitations of synthetic text.
  arXiv  Detail & Related papers  (2023-07-19T17:47:12Z)
• AugNet: End-to-End Unsupervised Visual Representation Learning with Image Augmentation [3.6790362352712873]
  We propose AugNet, a new deep learning training paradigm to learn image features from a collection of unlabeled pictures. Our experiments demonstrate that the method is able to represent the image in low dimensional space. Unlike many deep-learning-based image retrieval algorithms, our approach does not require access to external annotated datasets.
  arXiv  Detail & Related papers  (2021-06-11T09:02:30Z)
• Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision [57.031588264841]
  We leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme.
  arXiv  Detail & Related papers  (2021-02-11T10:08:12Z)
• Seed the Views: Hierarchical Semantic Alignment for Contrastive Representation Learning [116.91819311885166]
  We propose a hierarchical semantic alignment strategy via expanding the views generated by a single image to textbfCross-samples and Multi-level representation. Our method, termed as CsMl, has the ability to integrate multi-level visual representations across samples in a robust way.
  arXiv  Detail & Related papers  (2020-12-04T17:26:24Z)
• Fine-grained Visual Textual Alignment for Cross-Modal Retrieval using Transformer Encoders [14.634046503477979]
  We present a novel approach called Transformer Reasoning and Alignment Network (TERAN) TERAN enforces a fine-grained match between the underlying components of images and sentences. On the MS-COCO 1K test set, we obtain an improvement of 5.7% and 3.5% respectively on the image and the sentence retrieval tasks.
  arXiv  Detail & Related papers  (2020-08-12T11:02:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.