We introduce CLaMP: Contrastive Language-Music Pre-training, which learns cross-modal representations between natural language and symbolic music using a music encoder and a text encoder trained jointly with a contrastive loss. To pre-train CLaMP, we collected a large dataset of 1.4 million music-text pairs. It employed text dropout as a data augmentation technique and bar patching to efficiently represent music data which reduces sequence length to less than 10\%. In addition, we developed a masked music model pre-training objective to enhance the music encoder's comprehension of musical context and structure. CLaMP integrates textual information to enable semantic search and zero-shot classification for symbolic music, surpassing the capabilities of previous models. To support the evaluation of semantic search and music classification, we publicly release WikiMusicText (WikiMT), a dataset of 1010 lead sheets in ABC notation, each accompanied by a title, artist, genre, and description. In comparison to state-of-the-art models that require fine-tuning, zero-shot CLaMP demonstrated comparable or superior performance on score-oriented datasets. Our models and code are available at https://github.com/microsoft/muzic/tree/main/clamp.
Retrieving charts from a large corpus is a fundamental task that can benefit numerous applications such as visualization recommendations.The retrieved results are expected to conform to both explicit visual attributes (e.g., chart type, colormap) and implicit user intents (e.g., design style, context information) that vary upon application scenarios. However, existing example-based chart retrieval methods are built upon non-decoupled and low-level visual features that are hard to interpret, while definition-based ones are constrained to pre-defined attributes that are hard to extend. In this work, we propose a new framework, namely WYTIWYR (What-You-Think-Is-What-You-Retrieve), that integrates user intents into the chart retrieval process. The framework consists of two stages: first, the Annotation stage disentangles the visual attributes within the bitmap query chart; and second, the Retrieval stage embeds the user's intent with customized text prompt as well as query chart, to recall targeted retrieval result. We develop a prototype WYTIWYR system leveraging a contrastive language-image pre-training (CLIP) model to achieve zero-shot classification, and test the prototype on a large corpus with charts crawled from the Internet. Quantitative experiments, case studies, and qualitative interviews are conducted. The results demonstrate the usability and effectiveness of our proposed framework.
We present the first Africentric SemEval Shared task, Sentiment Analysis for African Languages (AfriSenti-SemEval) - The dataset is available at https://github.com/afrisenti-semeval/afrisent-semeval-2023. AfriSenti-SemEval is a sentiment classification challenge in 14 African languages: Amharic, Algerian Arabic, Hausa, Igbo, Kinyarwanda, Moroccan Arabic, Mozambican Portuguese, Nigerian Pidgin, Oromo, Swahili, Tigrinya, Twi, Xitsonga, and Yor\`ub\'a (Muhammad et al., 2023), using data labeled with 3 sentiment classes. We present three subtasks: (1) Task A: monolingual classification, which received 44 submissions; (2) Task B: multilingual classification, which received 32 submissions; and (3) Task C: zero-shot classification, which received 34 submissions. The best performance for tasks A and B was achieved by NLNDE team with 71.31 and 75.06 weighted F1, respectively. UCAS-IIE-NLP achieved the best average score for task C with 58.15 weighted F1. We describe the various approaches adopted by the top 10 systems and their approaches.
Large-scale Vision-Language Models, such as CLIP, learn powerful image-text representations that have found numerous applications, from zero-shot classification to text-to-image generation. Despite that, their capabilities for solving novel discriminative tasks via prompting fall behind those of large language models, such as GPT-3. Here we explore the idea of visual prompt engineering for solving computer vision tasks beyond classification by editing in image space instead of text. In particular, we discover an emergent ability of CLIP, where, by simply drawing a red circle around an object, we can direct the model's attention to that region, while also maintaining global information. We show the power of this simple approach by achieving state-of-the-art in zero-shot referring expressions comprehension and strong performance in keypoint localization tasks. Finally, we draw attention to some potential ethical concerns of large language-vision models.
We present RECLIP (Resource-efficient CLIP), a simple method that minimizes computational resource footprint for CLIP (Contrastive Language Image Pretraining). Inspired by the notion of coarse-to-fine in computer vision, we leverage small images to learn from large-scale language supervision efficiently, and finetune the model with high-resolution data in the end. Since the complexity of the vision transformer heavily depends on input image size, our approach significantly reduces the training resource requirements both in theory and in practice. Using the same batch size and training epoch, RECLIP achieves highly competitive zero-shot classification and image-text retrieval accuracy with 6 to 8x less computational resources and 7 to 9x fewer FLOPs than the baseline. Compared to the state-of-the-art contrastive learning methods, RECLIP demonstrates 5 to 59x training resource savings while maintaining highly competitive zero-shot classification and retrieval performance. Finally, RECLIP matches the state of the art in transfer learning to open-vocabulary detection tasks, achieving 32 APr on LVIS. We hope this work will pave the path for the broader research community to explore language supervised pretraining in resource-friendly settings.
Vision-Language models (VLMs) that use contrastive language-image pre-training have shown promising zero-shot classification performance. However, their performance on imbalanced dataset is relatively poor, where the distribution of classes in the training dataset is skewed, leading to poor performance in predicting minority classes. For instance, CLIP achieved only 5% accuracy on the iNaturalist18 dataset. We propose to add a lightweight decoder to VLMs to avoid OOM (out of memory) problem caused by large number of classes and capture nuanced features for tail classes. Then, we explore improvements of VLMs using prompt tuning, fine-tuning, and incorporating imbalanced algorithms such as Focal Loss, Balanced SoftMax and Distribution Alignment. Experiments demonstrate that the performance of VLMs can be further boosted when used with decoder and imbalanced methods. Specifically, our improved VLMs significantly outperforms zero-shot classification by an average accuracy of 6.58%, 69.82%, and 6.17%, on ImageNet-LT, iNaturalist18, and Places-LT, respectively. We further analyze the influence of pre-training data size, backbones, and training cost. Our study highlights the significance of imbalanced learning algorithms in face of VLMs pre-trained by huge data. We release our code at https://github.com/Imbalance-VLM/Imbalance-VLM.
During the preceding biennium, vision-language pre-training has achieved noteworthy success on several downstream tasks. Nevertheless, acquiring high-quality image-text pairs, where the pairs are entirely exclusive of each other, remains a challenging task, and noise exists in the commonly used datasets. To address this issue, we propose SoftCLIP, a novel approach that relaxes the strict one-to-one constraint and achieves a soft cross-modal alignment by introducing a softened target, which is generated from the fine-grained intra-modal self-similarity. The intra-modal guidance is indicative to enable two pairs have some local similarities and model many-to-many relationships between the two modalities. Besides, since the positive still dominates in the softened target distribution, we disentangle the negatives in the distribution to further boost the relation alignment with the negatives in the cross-modal learning. Extensive experiments demonstrate the effectiveness of SoftCLIP. In particular, on ImageNet zero-shot classification task, using CC3M/CC12M as pre-training dataset, SoftCLIP brings a top-1 accuracy improvement of 6.8%/7.2% over the CLIP baseline.
The recent wave of large-scale text-to-image diffusion models has dramatically increased our text-based image generation abilities. These models can generate realistic images for a staggering variety of prompts and exhibit impressive compositional generalization abilities. Almost all use cases thus far have solely focused on sampling; however, diffusion models can also provide conditional density estimates, which are useful for tasks beyond image generation. In this paper, we show that the density estimates from large-scale text-to-image diffusion models like Stable Diffusion can be leveraged to perform zero-shot classification without any additional training. Our generative approach to classification, which we call Diffusion Classifier, attains strong results on a variety of benchmarks and outperforms alternative methods of extracting knowledge from diffusion models. Although a gap remains between generative and discriminative approaches on zero-shot recognition tasks, our diffusion-based approach has significantly stronger multimodal compositional reasoning ability than competing discriminative approaches. Finally, we use Diffusion Classifier to extract standard classifiers from class-conditional diffusion models trained on ImageNet. Our models achieve strong classification performance using only weak augmentations and exhibit qualitatively better "effective robustness" to distribution shift. Overall, our results are a step toward using generative over discriminative models for downstream tasks. Results and visualizations at https://diffusion-classifier.github.io/
Large language models (LLM) have been successful in several natural language understanding tasks and could be relevant for natural language processing (NLP)-based mental health application research. In this work, we report the performance of LLM-based ChatGPT (with gpt-3.5-turbo backend) in three text-based mental health classification tasks: stress detection (2-class classification), depression detection (2-class classification), and suicidality detection (5-class classification). We obtained annotated social media posts for the three classification tasks from public datasets. Then ChatGPT API classified the social media posts with an input prompt for classification. We obtained F1 scores of 0.73, 0.86, and 0.37 for stress detection, depression detection, and suicidality detection, respectively. A baseline model that always predicted the dominant class resulted in F1 scores of 0.35, 0.60, and 0.19. The zero-shot classification accuracy obtained with ChatGPT indicates a potential use of language models for mental health classification tasks.
Prompt-OVD is an efficient and effective framework for open-vocabulary object detection that utilizes class embeddings from CLIP as prompts, guiding the Transformer decoder to detect objects in both base and novel classes. Additionally, our novel RoI-based masked attention and RoI pruning techniques help leverage the zero-shot classification ability of the Vision Transformer-based CLIP, resulting in improved detection performance at minimal computational cost. Our experiments on the OV-COCO and OVLVIS datasets demonstrate that Prompt-OVD achieves an impressive 21.2 times faster inference speed than the first end-to-end open-vocabulary detection method (OV-DETR), while also achieving higher APs than four two-stage-based methods operating within similar inference time ranges. Code will be made available soon.
The electrocardiogram (ECG) is one of the most commonly used non-invasive, convenient medical monitoring tools that assist in the clinical diagnosis of heart diseases. Recently, deep learning (DL) techniques, particularly self-supervised learning (SSL), have demonstrated great potential in the classification of ECG. SSL pre-training has achieved competitive performance with only a small amount of annotated data after fine-tuning. However, current SSL methods rely on the availability of annotated data and are unable to predict labels not existing in fine-tuning datasets. To address this challenge, we propose Multimodal ECG-Text Self-supervised pre-training (METS), the first work to utilize the auto-generated clinical reports to guide ECG SSL pre-training. We use a trainable ECG encoder and a frozen language model to embed paired ECG and automatically machine-generated clinical reports separately. The SSL aims to maximize the similarity between paired ECG and auto-generated report while minimize the similarity between ECG and other reports. In downstream classification tasks, METS achieves around 10% improvement in performance without using any annotated data via zero-shot classification, compared to other supervised and SSL baselines that rely on annotated data. Furthermore, METS achieves the highest recall and F1 scores on the MIT-BIH dataset, despite MIT-BIH containing different classes of ECG compared to the pre-trained dataset. The extensive experiments have demonstrated the advantages of using ECG-Text multimodal self-supervised learning in terms of generalizability, effectiveness, and efficiency.
This case study investigates the task of job classification in a real-world setting, where the goal is to determine whether an English-language job posting is appropriate for a graduate or entry-level position. We explore multiple approaches to text classification, including supervised approaches such as traditional models like Support Vector Machines (SVMs) and state-of-the-art deep learning methods such as DeBERTa. We compare them with Large Language Models (LLMs) used in both few-shot and zero-shot classification settings. To accomplish this task, we employ prompt engineering, a technique that involves designing prompts to guide the LLMs towards the desired output. Specifically, we evaluate the performance of two commercially available state-of-the-art GPT-3.5-based language models, text-davinci-003 and gpt-3.5-turbo. We also conduct a detailed analysis of the impact of different aspects of prompt engineering on the model's performance. Our results show that, with a well-designed prompt, a zero-shot gpt-3.5-turbo classifier outperforms all other models, achieving a 6% increase in Precision@95% Recall compared to the best supervised approach. Furthermore, we observe that the wording of the prompt is a critical factor in eliciting the appropriate "reasoning" in the model, and that seemingly minor aspects of the prompt significantly affect the model's performance.
Contrastive vision-language representation learning has achieved state-of-the-art performance for zero-shot classification, by learning from millions of image-caption pairs crawled from the internet. However, the massive data that powers large multimodal models such as CLIP, makes them extremely vulnerable to various types of targeted data poisoning and backdoor attacks. Despite this vulnerability, robust contrastive vision-language pre-training against such attacks has remained unaddressed. In this work, we propose ROCLIP, the first effective method for robust pre-training multimodal vision-language models against targeted data poisoning and backdoor attacks. ROCLIP effectively breaks the association between poisoned image-caption pairs by considering a relatively large and varying pool of random captions, and matching every image with the text that is most similar to it in the pool instead of its own caption, every few epochs.It also leverages image and text augmentations to further strengthen the defense and improve the performance of the model. Our extensive experiments show that ROCLIP renders state-of-the-art targeted data poisoning and backdoor attacks ineffective during pre-training CLIP models. In particular, ROCLIP decreases the success rate for targeted data poisoning attacks from 93.75% to 12.5% and that of backdoor attacks down to 0%, while improving the model's linear probe performance by 10% and maintains a similar zero shot performance compared to CLIP. By increasing the frequency of matching, ROCLIP is able to defend strong attacks, which add up to 1% poisoned examples to the data, and successfully maintain a low attack success rate of 12.5%, while trading off the performance on some tasks.
Text-based Person Search (TPS), is targeted on retrieving pedestrians to match text descriptions instead of query images. Recent Vision-Language Pre-training (VLP) models can bring transferable knowledge to downstream TPS tasks, resulting in more efficient performance gains. However, existing TPS methods improved by VLP only utilize pre-trained visual encoders, neglecting the corresponding textual representation and breaking the significant modality alignment learned from large-scale pre-training. In this paper, we explore the full utilization of textual potential from VLP in TPS tasks. We build on the proposed VLP-TPS baseline model, which is the first TPS model with both pre-trained modalities. We propose the Multi-Integrity Description Constraints (MIDC) to enhance the robustness of the textual modality by incorporating different components of fine-grained corpus during training. Inspired by the prompt approach for zero-shot classification with VLP models, we propose the Dynamic Attribute Prompt (DAP) to provide a unified corpus of fine-grained attributes as language hints for the image modality. Extensive experiments show that our proposed TPS framework achieves state-of-the-art performance, exceeding the previous best method by a margin.
Human speakers can generate descriptions of perceptual concepts, abstracted from the instance-level. Moreover, such descriptions can be used by other speakers to learn provisional representations of those concepts. Learning and using abstract perceptual concepts is under-investigated in the language-and-vision field. The problem is also highly relevant to the field of representation learning in multi-modal NLP. In this paper, we introduce a framework for testing category-level perceptual grounding in multi-modal language models. In particular, we train separate neural networks to generate and interpret descriptions of visual categories. We measure the communicative success of the two models with the zero-shot classification performance of the interpretation model, which we argue is an indicator of perceptual grounding. Using this framework, we compare the performance of prototype- and exemplar-based representations. Finally, we show that communicative success exposes performance issues in the generation model, not captured by traditional intrinsic NLG evaluation metrics, and argue that these issues stem from a failure to properly ground language in vision at the category level.
Large vision-language models (VLMs), such as CLIP, learn rich joint image-text representations, facilitating advances in numerous downstream tasks, including zero-shot classification and text-to-image generation. Nevertheless, existing VLMs exhibit a prominent well-documented limitation - they fail to encapsulate compositional concepts such as counting. We introduce a simple yet effective method to improve the quantitative understanding of VLMs, while maintaining their overall performance on common benchmarks. Specifically, we propose a new counting-contrastive loss used to finetune a pre-trained VLM in tandem with its original objective. Our counting loss is deployed over automatically-created counterfactual examples, each consisting of an image and a caption containing an incorrect object count. For example, an image depicting three dogs is paired with the caption "Six dogs playing in the yard". Our loss encourages discrimination between the correct caption and its counterfactual variant which serves as a hard negative example. To the best of our knowledge, this work is the first to extend CLIP's capabilities to object counting. Furthermore, we introduce "CountBench" - a new image-text counting benchmark for evaluating a model's understanding of object counting. We demonstrate a significant improvement over state-of-the-art baseline models on this task. Finally, we leverage our count-aware CLIP model for image retrieval and text-conditioned image generation, demonstrating that our model can produce specific counts of objects more reliably than existing ones.
Recent advances in training vision-language models have demonstrated unprecedented robustness and transfer learning effectiveness; however, standard computer vision datasets are image-only, and therefore not well adapted to such training methods. Our paper introduces a simple methodology for adapting any fine-grained image classification dataset for distributed vision-language pretraining. We implement this methodology on the challenging iNaturalist-2021 dataset, comprised of approximately 2.7 million images of macro-organisms across 10,000 classes, and achieve a new state-of-the art model in terms of zero-shot classification accuracy. Somewhat surprisingly, our model (trained using a new method called locked-image text tuning) uses a pre-trained, frozen vision representation, proving that language alignment alone can attain strong transfer learning performance, even on fractious, long-tailed datasets. Our approach opens the door for utilizing high quality vision-language pretrained models in agriculturally relevant applications involving species detection.
Transferring learned patterns from pretrained neural language models has been shown to significantly improve effectiveness across a variety of language-based tasks, meanwhile further tuning on intermediate tasks has been demonstrated to provide additional performance benefits, provided the intermediate task is sufficiently related to the target task. However, how to identify related tasks is an open problem, and brute-force searching effective task combinations is prohibitively expensive. Hence, the question arises, are we able to improve the effectiveness and efficiency of tasks with no training examples through selective fine-tuning? In this paper, we explore statistical measures that approximate the divergence between domain representations as a means to estimate whether tuning using one task pair will exhibit performance benefits over tuning another. This estimation can then be used to reduce the number of task pairs that need to be tested by eliminating pairs that are unlikely to provide benefits. Through experimentation over 58 tasks and over 6,600 task pair combinations, we demonstrate that statistical measures can distinguish effective task pairs, and the resulting estimates can reduce end-to-end runtime by up to 40%.
Contrastively trained text-image models have the remarkable ability to perform zero-shot classification, that is, classifying previously unseen images into categories that the model has never been explicitly trained to identify. However, these zero-shot classifiers need prompt engineering to achieve high accuracy. Prompt engineering typically requires hand-crafting a set of prompts for individual downstream tasks. In this work, we aim to automate this prompt engineering and improve zero-shot accuracy through prompt ensembling. In particular, we ask "Given a large pool of prompts, can we automatically score the prompts and ensemble those that are most suitable for a particular downstream dataset, without needing access to labeled validation data?". We demonstrate that this is possible. In doing so, we identify several pathologies in a naive prompt scoring method where the score can be easily overconfident due to biases in pre-training and test data, and we propose a novel prompt scoring method that corrects for the biases. Using our proposed scoring method to create a weighted average prompt ensemble, our method outperforms equal average ensemble, as well as hand-crafted prompts, on ImageNet, 4 of its variants, and 11 fine-grained classification benchmarks, all while being fully automatic, optimization-free, and not requiring access to labeled validation data.
In this work, we investigate the problem of Model-Agnostic Zero-Shot Classification (MA-ZSC), which refers to training non-specific classification architectures (downstream models) to classify real images without using any real images during training. Recent research has demonstrated that generating synthetic training images using diffusion models provides a potential solution to address MA-ZSC. However, the performance of this approach currently falls short of that achieved by large-scale vision-language models. One possible explanation is a potential significant domain gap between synthetic and real images. Our work offers a fresh perspective on the problem by providing initial insights that MA-ZSC performance can be improved by improving the diversity of images in the generated dataset. We propose a set of modifications to the text-to-image generation process using a pre-trained diffusion model to enhance diversity, which we refer to as our $\textbf{bag of tricks}$. Our approach shows notable improvements in various classification architectures, with results comparable to state-of-the-art models such as CLIP. To validate our approach, we conduct experiments on CIFAR10, CIFAR100, and EuroSAT, which is particularly difficult for zero-shot classification due to its satellite image domain. We evaluate our approach with five classification architectures, including ResNet and ViT. Our findings provide initial insights into the problem of MA-ZSC using diffusion models. All code will be available on GitHub.