Severe data imbalance naturally exists among web-scale vision-language datasets. Despite this, we find CLIP pre-trained thereupon exhibits notable robustness to the data imbalance compared to supervised learning, and demonstrates significant effectiveness in learning generalizable representations. With an aim to investigate the reasons behind this finding, we conduct controlled experiments to study various underlying factors, and reveal that CLIP's pretext task forms a dynamic classification problem wherein only a subset of classes is present in training. This isolates the bias from dominant classes and implicitly balances the learning signal. Furthermore, the robustness and discriminability of CLIP improve with more descriptive language supervision, larger data scale, and broader open-world concepts, which are inaccessible to supervised learning. Our study not only uncovers the mechanisms behind CLIP's generalizability beyond data imbalance but also provides transferable insights for the research community. The findings are validated in both supervised and self-supervised learning, enabling models trained on imbalanced data to achieve CLIP-level performance on diverse recognition tasks. Code and data are available at: https://github.com/CVMI-Lab/clip-beyond-tail.
Semantic channel equalization has emerged as a solution to address language mismatch in multi-user semantic communications. This approach aims to align the latent spaces of an encoder and a decoder which were not jointly trained and it relies on a partition of the semantic (latent) space into atoms based on the the semantic meaning. In this work we explore the role of the semantic space partition in scenarios where the task structure involves a one-to-many mapping between the semantic space and the action space. In such scenarios, partitioning based on hard inference results results in loss of information which degrades the equalization performance. We propose a soft criterion to derive the atoms of the partition which leverages the soft decoder's output and offers a more comprehensive understanding of the semantic space's structure. Through empirical validation, we demonstrate that soft partitioning yields a more descriptive and regular partition of the space, consequently enhancing the performance of the equalization algorithm.
In this paper, we present an approach in the Multimodal Learning Analytics field. Within this approach, we have developed a tool to visualize and analyze eye movement data collected during learning sessions in online courses. The tool is named VAAD, an acronym for Visual Attention Analysis Dashboard. These eye movement data have been gathered using an eye-tracker and subsequently processed and visualized for interpretation. The purpose of the tool is to conduct a descriptive analysis of the data by facilitating its visualization, enabling the identification of differences and learning patterns among various learner populations. Additionally, it integrates a predictive module capable of anticipating learner activities during a learning session. Consequently, VAAD holds the potential to offer valuable insights into online learning behaviors from both descriptive and predictive perspectives.
A key challenge for the widespread application of learning-based models for robotic perception is to significantly reduce the required amount of annotated training data while achieving accurate predictions. This is essential not only to decrease operating costs but also to speed up deployment time. In this work, we address this challenge for PAnoptic SegmenTation with fEw Labels (PASTEL) by exploiting the groundwork paved by visual foundation models. We leverage descriptive image features from such a model to train two lightweight network heads for semantic segmentation and object boundary detection, using very few annotated training samples. We then merge their predictions via a novel fusion module that yields panoptic maps based on normalized cut. To further enhance the performance, we utilize self-training on unlabeled images selected by a feature-driven similarity scheme. We underline the relevance of our approach by employing PASTEL to important robot perception use cases from autonomous driving and agricultural robotics. In extensive experiments, we demonstrate that PASTEL significantly outperforms previous methods for label-efficient segmentation even when using fewer annotations. The code of our work is publicly available at http://pastel.cs.uni-freiburg.de.
With the rapid advancement of Vision Language Models (VLMs), VLM-based Image Quality Assessment (IQA) seeks to describe image quality linguistically to align with human expression and capture the multifaceted nature of IQA tasks. However, current methods are still far from practical usage. First, prior works focus narrowly on specific sub-tasks or settings, which do not align with diverse real-world applications. Second, their performance is sub-optimal due to limitations in dataset coverage, scale, and quality. To overcome these challenges, we introduce Depicted image Quality Assessment in the Wild (DepictQA-Wild). Our method includes a multi-functional IQA task paradigm that encompasses both assessment and comparison tasks, brief and detailed responses, full-reference and non-reference scenarios. We introduce a ground-truth-informed dataset construction approach to enhance data quality, and scale up the dataset to 495K under the brief-detail joint framework. Consequently, we construct a comprehensive, large-scale, and high-quality dataset, named DQ-495K. We also retain image resolution during training to better handle resolution-related quality issues, and estimate a confidence score that is helpful to filter out low-quality responses. Experimental results demonstrate that DepictQA-Wild significantly outperforms traditional score-based methods, prior VLM-based IQA models, and proprietary GPT-4V in distortion identification, instant rating, and reasoning tasks. Our advantages are further confirmed by real-world applications including assessing the web-downloaded images and ranking model-processed images. Datasets and codes will be released in https://depictqa.github.io/depictqa-wild/.
Text-to-image generation models have seen considerable advancement, catering to the increasing interest in personalized image creation. Current customization techniques often necessitate users to provide multiple images (typically 3-5) for each customized object, along with the classification of these objects and descriptive textual prompts for scenes. This paper questions whether the process can be made more user-friendly and the customization more intricate. We propose a method where users need only provide images along with text for each customization topic, and necessitates only a single image per visual concept. We introduce the concept of a ``multi-modal prompt'', a novel integration of text and images tailored to each customization concept, which simplifies user interaction and facilitates precise customization of both objects and scenes. Our proposed paradigm for customized text-to-image generation surpasses existing finetune-based methods in user-friendliness and the ability to customize complex objects with user-friendly inputs. Our code is available at $\href{https://github.com/zhongzero/Multi-Modal-Prompt}{https://github.com/zhongzero/Multi-Modal-Prompt}$.
Due to the need to interact with the real world, embodied agents are required to possess comprehensive prior knowledge, long-horizon planning capability, and a swift response speed. Despite recent large language model (LLM) based agents achieving promising performance, they still exhibit several limitations. For instance, the output of LLMs is a descriptive sentence, which is ambiguous when determining specific actions. To address these limitations, we introduce the large auto-regressive model (LARM). LARM leverages both text and multi-view images as input and predicts subsequent actions in an auto-regressive manner. To train LARM, we develop a novel data format named auto-regressive node transmission structure and assemble a corresponding dataset. Adopting a two-phase training regimen, LARM successfully harvests enchanted equipment in Minecraft, which demands significantly more complex decision-making chains than the highest achievements of prior best methods. Besides, the speed of LARM is 6.8x faster.
This work introduces composed image retrieval to remote sensing. It allows to query a large image archive by image examples alternated by a textual description, enriching the descriptive power over unimodal queries, either visual or textual. Various attributes can be modified by the textual part, such as shape, color, or context. A novel method fusing image-to-image and text-to-image similarity is introduced. We demonstrate that a vision-language model possesses sufficient descriptive power and no further learning step or training data are necessary. We present a new evaluation benchmark focused on color, context, density, existence, quantity, and shape modifications. Our work not only sets the state-of-the-art for this task, but also serves as a foundational step in addressing a gap in the field of remote sensing image retrieval. Code at: https://github.com/billpsomas/rscir
Training neural networks with high-quality pixel-level annotation in histopathology whole-slide images (WSI) is an expensive process due to gigapixel resolution of WSIs. However, recent advances in self-supervised learning have shown that highly descriptive image representations can be learned without the need for annotations. We investigate the application of the recent Hierarchical Image Pyramid Transformer (HIPT) model for the specific task of classification of colorectal biopsies and polyps. After evaluating the effectiveness of TCGA-learned features in the original HIPT model, we incorporate colon biopsy image information into HIPT's pretraining using two distinct strategies: (1) fine-tuning HIPT from the existing TCGA weights and (2) pretraining HIPT from random weight initialization. We compare the performance of these pretraining regimes on two colorectal biopsy classification tasks: binary and multiclass classification.
The absence of adequately sufficient expert-level tumor annotations hinders the effectiveness of supervised learning based opportunistic cancer screening on medical imaging. Clinical reports (that are rich in descriptive textual details) can offer a "free lunch'' supervision information and provide tumor location as a type of weak label to cope with screening tasks, thus saving human labeling workloads, if properly leveraged. However, predicting cancer only using such weak labels can be very changeling since tumors are usually presented in small anatomical regions compared to the whole 3D medical scans. Weakly semi-supervised learning (WSSL) utilizes a limited set of voxel-level tumor annotations and incorporates alongside a substantial number of medical images that have only off-the-shelf clinical reports, which may strike a good balance between minimizing expert annotation workload and optimizing screening efficacy. In this paper, we propose a novel text-guided learning method to achieve highly accurate cancer detection results. Through integrating diagnostic and tumor location text prompts into the text encoder of a vision-language model (VLM), optimization of weakly supervised learning can be effectively performed in the latent space of VLM, thereby enhancing the stability of training. Our approach can leverage clinical knowledge by large-scale pre-trained VLM to enhance generalization ability, and produce reliable pseudo tumor masks to improve cancer detection. Our extensive quantitative experimental results on a large-scale cancer dataset, including 1,651 unique patients, validate that our approach can reduce human annotation efforts by at least 70% while maintaining comparable cancer detection accuracy to competing fully supervised methods (AUC value 0.961 versus 0.966).
This paper presents a highly speculative model encompassing the cortex, thalamus, and hippocampus of the mammalian brain. While the majority of computational neuroscience models are founded upon empirical evidence, this model is predicated upon a hardware proposal for a machine learning accelerator. Such a device was designed to perform a specific task, such as speech recognition. The design process employed the principles and techniques typically used by computer architects in the design of devices such as processors. However, it also sought to maintain plausibility with biological systems in accordance with the current understanding of the mammalian brain. In the course of our research, we have identified a functional framework that may help to fill the gaps in current neuroscience, thereby facilitating the explanations for many elusive cognitive-level effects. This paper does not describe the device itself or the rationale behind the design decision, but instead, it presents a concise description of the derived model. In brief, the model provides a functional definition of the cortical column and its structural definition by the minicolumns. It also offers a descriptive model for the corticothalamic and corticostriatal loops, a functional proposal for the hippocampal complex, and a simplified view of the brainstem circuitry involved in auditory processing. The proposed model appears to provide an explanation for a number of cognitive phenomena, including some ERP effects, bottom-up and top-down attention, and the relationship between phenomena such as the cocktail party effect, anterograde and retrograde amnesia following hippocampal complex damage, and so forth.
In some markets, the visual appearance of a product matters a lot. This paper investigates consumer transactions from a major fashion retailer, focusing on consumer aesthetics. Pretrained multimodal models convert images and text descriptions into high-dimensional embeddings. The value of these embeddings is verified both empirically and by their ability to segment the product space. A discrete choice model is used to decompose the distinct drivers of consumer choice: price, visual aesthetics, descriptive details, and seasonal variations. Consumers are allowed to differ in their preferences over these factors, both through observed variation in demographics and allowing for unobserved types. Estimation and inference employ automatic differentiation and GPUs, making it scalable and portable. The model reveals significant differences in price sensitivity and aesthetic preferences across consumers. The model is validated by its ability to predict the relative success of new designs and purchase patterns.
The remarkable progress of Multi-modal Large Language Models (MLLMs) has attracted significant attention due to their superior performance in visual contexts. However, their capabilities in turning visual figure to executable code, have not been evaluated thoroughly. To address this, we introduce Plot2Code, a comprehensive visual coding benchmark designed for a fair and in-depth assessment of MLLMs. We carefully collect 132 manually selected high-quality matplotlib plots across six plot types from publicly available matplotlib galleries. For each plot, we carefully offer its source code, and an descriptive instruction summarized by GPT-4. This approach enables Plot2Code to extensively evaluate MLLMs' code capabilities across various input modalities. Furthermore, we propose three automatic evaluation metrics, including code pass rate, text-match ratio, and GPT-4V overall rating, for a fine-grained assessment of the output code and rendered images. Instead of simply judging pass or fail, we employ GPT-4V to make an overall judgement between the generated and reference images, which has been shown to be consistent with human evaluation. The evaluation results, which include analyses of 14 MLLMs such as the proprietary GPT-4V, Gemini-Pro, and the open-sourced Mini-Gemini, highlight the substantial challenges presented by Plot2Code. With Plot2Code, we reveal that most existing MLLMs struggle with visual coding for text-dense plots, heavily relying on textual instruction. We hope that the evaluation results from Plot2Code on visual coding will guide the future development of MLLMs. All data involved with Plot2Code are available at https://huggingface.co/datasets/TencentARC/Plot2Code.
Phishing email attacks are among the most common and most harmful cybersecurity attacks. With the emergence of generative AI, phishing attacks can be based on emails generated automatically, making it more difficult to detect them. That is, instead of a single email format sent to a large number of recipients, generative AI can be used to send each potential victim a different email, making it more difficult for cybersecurity systems to identify the scam email before it reaches the recipient. Here we describe a corpus of AI-generated phishing emails. We also use different machine learning tools to test the ability of automatic text analysis to identify AI-generated phishing emails. The results are encouraging, and show that machine learning tools can identify an AI-generated phishing email with high accuracy compared to regular emails or human-generated scam email. By applying descriptive analytic, the specific differences between AI-generated emails and manually crafted scam emails are profiled, and show that AI-generated emails are different in their style from human-generated phishing email scams. Therefore, automatic identification tools can be used as a warning for the user. The paper also describes the corpus of AI-generated phishing emails that is made open to the public, and can be used for consequent studies. While the ability of machine learning to detect AI-generated phishing email is encouraging, AI-generated phishing emails are different from regular phishing emails, and therefore it is important to train machine learning systems also with AI-generated emails in order to repel future phishing attacks that are powered by generative AI.
I explored adapting Stable Diffusion v1.5 for generating domain-specific satellite and aerial images in remote sensing. Recognizing the limitations of existing models like Midjourney and Stable Diffusion, trained primarily on natural RGB images and lacking context for remote sensing, I used the RSICD dataset to train a Stable Diffusion model with a loss of 0.2. I incorporated descriptive captions from the dataset for text-conditioning. Additionally, I created a synthetic dataset for a Land Use Land Classification (LULC) task, employing prompting techniques with RAG and ChatGPT and fine-tuning a specialized remote sensing LLM. However, I faced challenges with prompt quality and model performance. I trained a classification model (ResNet18) on the synthetic dataset achieving 49.48% test accuracy in TorchGeo to create a baseline. Quantitative evaluation through FID scores and qualitative feedback from domain experts assessed the realism and quality of the generated images and dataset. Despite extensive fine-tuning and dataset iterations, results indicated subpar image quality and realism, as indicated by high FID scores and domain-expert evaluation. These findings call attention to the potential of diffusion models in remote sensing while highlighting significant challenges related to insufficient pretraining data and computational resources.
Numerous sign language datasets exist, yet they typically cover only a limited selection of the thousands of signs used globally. Moreover, creating diverse sign language datasets is an expensive and challenging task due to the costs associated with gathering a varied group of signers. Motivated by these challenges, we aimed to develop a solution that addresses these limitations. In this context, we focused on textually describing body movements from skeleton keypoint sequences, leading to the creation of a new dataset. We structured this dataset around AUTSL, a comprehensive isolated Turkish sign language dataset. We also developed a baseline model, SkelCap, which can generate textual descriptions of body movements. This model processes the skeleton keypoints data as a vector, applies a fully connected layer for embedding, and utilizes a transformer neural network for sequence-to-sequence modeling. We conducted extensive evaluations of our model, including signer-agnostic and sign-agnostic assessments. The model achieved promising results, with a ROUGE-L score of 0.98 and a BLEU-4 score of 0.94 in the signer-agnostic evaluation. The dataset we have prepared, namely the AUTSL-SkelCap, will be made publicly available soon.
In this paper, a time series algorithm based on Genetic Algorithm (GA) and Long Short-Term Memory Network (LSTM) optimization is used to forecast stock prices effectively, taking into account the trend of the big data era. The data are first analyzed by descriptive statistics, and then the model is built and trained and tested on the dataset. After optimization and adjustment, the mean absolute error (MAE) of the model gradually decreases from 0.11 to 0.01 and tends to be stable, indicating that the model prediction effect is gradually close to the real value. The results on the test set show that the time series algorithm optimized based on Genetic Algorithm (GA)-Long Short-Term Memory Network (LSTM) is able to accurately predict the stock prices, and is highly consistent with the actual price trends and values, with strong generalization ability. The MAE on the test set is 2.41, the MSE is 9.84, the RMSE is 3.13, and the R2 is 0.87. This research result not only provides a novel stock price prediction method, but also provides a useful reference for financial market analysis using computer technology and big data.
In recent years, AI-Generated Content (AIGC) has witnessed rapid advancements, facilitating the creation of music, images, and other artistic forms across a wide range of industries. However, current models for image- and video-to-music synthesis struggle to capture the nuanced emotions and atmosphere conveyed by visual content. To fill this gap, we propose Mozart's Touch, a multi-modal music generation framework capable of generating music aligned with cross-modal inputs such as images, videos, and text. The framework consists of three key components: Multi-modal Captioning Module, Large Language Model (LLM) understanding \& Bridging Module, and Music Generation Module. Unlike traditional end-to-end methods, Mozart's Touch uses LLMs to accurately interpret visual elements without requiring the training or fine-tuning of music generation models, providing efficiency and transparency through clear, interpretable prompts. We also introduce the "LLM-Bridge" method to resolve the heterogeneous representation challenges between descriptive texts from different modalities. Through a series of objective and subjective evaluations, we demonstrate that Mozart's Touch outperforms current state-of-the-art models. Our code and examples are available at https://github.com/TiffanyBlews/MozartsTouch.
Radiology report generation aims to automatically generate detailed and coherent descriptive reports alongside radiology images. Previous work mainly focused on refining fine-grained image features or leveraging external knowledge. However, the precise alignment of fine-grained image features with corresponding text descriptions has not been considered. This paper presents a novel method called Fine-grained Image-Text Aligner (FITA) to construct fine-grained alignment for image and text features. It has three novel designs: Image Feature Refiner (IFR), Text Feature Refiner (TFR) and Contrastive Aligner (CA). IFR and TFR aim to learn fine-grained image and text features, respectively. We achieve this by leveraging saliency maps to effectively fuse symptoms with corresponding abnormal visual regions, and by utilizing a meticulously constructed triplet set for training. Finally, CA module aligns fine-grained image and text features using contrastive loss for precise alignment. Results show that our method surpasses existing methods on the widely used benchmark
This paper introduces CookingSense, a descriptive collection of knowledge assertions in the culinary domain extracted from various sources, including web data, scientific papers, and recipes, from which knowledge covering a broad range of aspects is acquired. CookingSense is constructed through a series of dictionary-based filtering and language model-based semantic filtering techniques, which results in a rich knowledgebase of multidisciplinary food-related assertions. Additionally, we present FoodBench, a novel benchmark to evaluate culinary decision support systems. From evaluations with FoodBench, we empirically prove that CookingSense improves the performance of retrieval augmented language models. We also validate the quality and variety of assertions in CookingSense through qualitative analysis.