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.
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 $\href{https://github.com/zhongzero/Multi-Modal-Prompt}{https://github.com/zhongzero/Multi-Modal-Prompt}$.
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
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.
In high-pressure environments where human individuals must simultaneously monitor multiple entities, communicate effectively, and maintain intense focus, the perception of time becomes a critical factor influencing performance and well-being. One indicator of well-being can be the person's subjective time perception. In our project $ChronoPilot$, we aim to develop a device that modulates human subjective time perception. In this study, we present a method to automatically assess the subjective time perception of air traffic controllers, a group often faced with demanding conditions, using their physiological data and eleven state-of-the-art machine learning classifiers. The physiological data consist of photoplethysmogram, electrodermal activity, and temperature data. We find that the support vector classifier works best with an accuracy of 79 % and electrodermal activity provides the most descriptive biomarker. These findings are an important step towards closing the feedback loop of our $ChronoPilot$-device to automatically modulate the user's subjective time perception. This technological advancement may promise improvements in task management, stress reduction, and overall productivity in high-stakes professions.
We study interactive learning of LLM-based language agents based on user edits made to the agent's output. In a typical setting such as writing assistants, the user interacts with a language agent to generate a response given a context, and may optionally edit the agent response to personalize it based on their latent preference, in addition to improving the correctness. The edit feedback is naturally generated, making it a suitable candidate for improving the agent's alignment with the user's preference, and for reducing the cost of user edits over time. We propose a learning framework, PRELUDE that infers a description of the user's latent preference based on historic edit data. The inferred user preference descriptions are used to define prompts for generating responses in the future. This avoids fine-tuning the agent, which is costly, challenging to scale with the number of users, and may even degrade its performance on other tasks. Furthermore, learning descriptive preference improves interpretability, allowing the user to view and modify the learned preference. However, user preference can be complex, subtle, and vary based on context, making it challenging to learn. To address this, we propose a simple yet effective algorithm named CIPHER that leverages the LLM to infer the user preference for a given context based on user edits. In the future, CIPHER retrieves inferred preferences from the k-closest contexts in the history, and forms an aggregate preference for response generation. We introduce two interactive environments -- summarization and email writing, and use a GPT-4 simulated user for evaluation. On both tasks, CIPHER outperforms several baselines by achieving the lowest edit distance cost while only having a small overhead in LLM query cost. Our analysis reports that user preferences learned by CIPHER show significant similarity to the ground truth latent preferences.
Text-to-Image (T2I) Synthesis has made tremendous strides in enhancing synthesized image quality, but current datasets evaluate model performance only on descriptive, instruction-based prompts. Real-world news image captions take a more pragmatic approach, providing high-level situational and Named-Entity (NE) information and limited physical object descriptions, making them abstractive. To evaluate the ability of T2I models to capture intended subjects from news captions, we introduce the Abstractive News Captions with High-level cOntext Representation (ANCHOR) dataset, containing 70K+ samples sourced from 5 different news media organizations. With Large Language Models (LLM) achieving success in language and commonsense reasoning tasks, we explore the ability of different LLMs to identify and understand key subjects from abstractive captions. Our proposed method Subject-Aware Finetuning (SAFE), selects and enhances the representation of key subjects in synthesized images by leveraging LLM-generated subject weights. It also adapts to the domain distribution of news images and captions through custom Domain Fine-tuning, outperforming current T2I baselines on ANCHOR. By launching the ANCHOR dataset, we hope to motivate research in furthering the Natural Language Understanding (NLU) capabilities of T2I models.
Multi-modal knowledge graph completion (MMKGC) aims to discover unobserved knowledge from given knowledge graphs, collaboratively leveraging structural information from the triples and multi-modal information of the entities to overcome the inherent incompleteness. Existing MMKGC methods usually extract multi-modal features with pre-trained models, resulting in coarse handling of multi-modal entity information, overlooking the nuanced, fine-grained semantic details and their complex interactions. To tackle this shortfall, we introduce a novel framework MyGO to tokenize, fuse, and augment the fine-grained multi-modal representations of entities and enhance the MMKGC performance. Motivated by the tokenization technology, MyGO tokenizes multi-modal entity information as fine-grained discrete tokens and learns entity representations with a cross-modal entity encoder. To further augment the multi-modal representations, MyGO incorporates fine-grained contrastive learning to highlight the specificity of the entity representations. Experiments on standard MMKGC benchmarks reveal that our method surpasses 19 of the latest models, underlining its superior performance. Code and data can be found in https://github.com/zjukg/MyGO
Traffic video description and analysis have received much attention recently due to the growing demand for efficient and reliable urban surveillance systems. Most existing methods only focus on locating traffic event segments, which severely lack descriptive details related to the behaviour and context of all the subjects of interest in the events. In this paper, we present TrafficVLM, a novel multi-modal dense video captioning model for vehicle ego camera view. TrafficVLM models traffic video events at different levels of analysis, both spatially and temporally, and generates long fine-grained descriptions for the vehicle and pedestrian at different phases of the event. We also propose a conditional component for TrafficVLM to control the generation outputs and a multi-task fine-tuning paradigm to enhance TrafficVLM's learning capability. Experiments show that TrafficVLM performs well on both vehicle and overhead camera views. Our solution achieved outstanding results in Track 2 of the AI City Challenge 2024, ranking us third in the challenge standings. Our code is publicly available at https://github.com/quangminhdinh/TrafficVLM.
Concept Bottleneck Models (CBMs) map the black-box visual representations extracted by deep neural networks onto a set of interpretable concepts and use the concepts to make predictions, enhancing the transparency of the decision-making process. Multimodal pre-trained models can match visual representations with textual concept embeddings, allowing for obtaining the interpretable concept bottleneck without the expertise concept annotations. Recent research has focused on the concept bank establishment and the high-quality concept selection. However, it is challenging to construct a comprehensive concept bank through humans or large language models, which severely limits the performance of CBMs. In this work, we propose the Incremental Residual Concept Bottleneck Model (Res-CBM) to address the challenge of concept completeness. Specifically, the residual concept bottleneck model employs a set of optimizable vectors to complete missing concepts, then the incremental concept discovery module converts the complemented vectors with unclear meanings into potential concepts in the candidate concept bank. Our approach can be applied to any user-defined concept bank, as a post-hoc processing method to enhance the performance of any CBMs. Furthermore, to measure the descriptive efficiency of CBMs, the Concept Utilization Efficiency (CUE) metric is proposed. Experiments show that the Res-CBM outperforms the current state-of-the-art methods in terms of both accuracy and efficiency and achieves comparable performance to black-box models across multiple datasets.
Generative models have shown a giant leap in synthesizing photo-realistic images with minimal expertise, sparking concerns about the authenticity of online information. This study aims to develop a universal AI-generated image detector capable of identifying images from diverse sources. Existing methods struggle to generalize across unseen generative models when provided with limited sample sources. Inspired by the zero-shot transferability of pre-trained vision-language models, we seek to harness the nontrivial visual-world knowledge and descriptive proficiency of CLIP-ViT to generalize over unknown domains. This paper presents a novel parameter-efficient fine-tuning approach, mixture of low-rank experts, to fully exploit CLIP-ViT's potential while preserving knowledge and expanding capacity for transferable detection. We adapt only the MLP layers of deeper ViT blocks via an integration of shared and separate LoRAs within an MoE-based structure. Extensive experiments on public benchmarks show that our method achieves superiority over state-of-the-art approaches in cross-generator generalization and robustness to perturbations. Remarkably, our best-performing ViT-L/14 variant requires training only 0.08% of its parameters to surpass the leading baseline by +3.64% mAP and +12.72% avg.Acc across unseen diffusion and autoregressive models. This even outperforms the baseline with just 0.28% of the training data. Our code and pre-trained models will be available at https://github.com/zhliuworks/CLIPMoLE.
We propose VecKM, a local point cloud geometry encoder that is descriptive and efficient to compute. VecKM leverages a unique approach by vectorizing a kernel mixture to represent the local point cloud. Such representation's descriptiveness is supported by two theorems that validate its ability to reconstruct and preserve the similarity of the local shape. Unlike existing encoders downsampling the local point cloud, VecKM constructs the local geometry encoding using all neighboring points, producing a more descriptive encoding. Moreover, VecKM is efficient to compute and scalable to large point cloud inputs: VecKM reduces the memory cost from $(n^2+nKd)$ to $(nd+np)$; and reduces the major runtime cost from computing $nK$ MLPs to $n$ MLPs, where $n$ is the size of the point cloud, $K$ is the neighborhood size, $d$ is the encoding dimension, and $p$ is a marginal factor. The efficiency is due to VecKM's unique factorizable property that eliminates the need of explicitly grouping points into neighbors. In the normal estimation task, VecKM demonstrates not only 100x faster inference speed but also highest accuracy and strongest robustness. In classification and segmentation tasks, integrating VecKM as a preprocessing module achieves consistently better performance than the PointNet, PointNet++, and point transformer baselines, and runs consistently faster by up to 10 times.
Descriptive grammars are highly valuable, but writing them is time-consuming and difficult. Furthermore, while linguists typically use corpora to create them, grammar descriptions often lack quantitative data. As for formal grammars, they can be challenging to interpret. In this paper, we propose a new method to extract and explore significant fine-grained grammar patterns and potential syntactic grammar rules from treebanks, in order to create an easy-to-understand corpus-based grammar. More specifically, we extract descriptions and rules across different languages for two linguistic phenomena, agreement and word order, using a large search space and paying special attention to the ranking order of the extracted rules. For that, we use a linear classifier to extract the most salient features that predict the linguistic phenomena under study. We associate statistical information to each rule, and we compare the ranking of the model's results to those of other quantitative and statistical measures. Our method captures both well-known and less well-known significant grammar rules in Spanish, French, and Wolof.
Recent works in Task and Motion Planning (TAMP) show that training control policies on language-supervised robot trajectories with quality labeled data markedly improves agent task success rates. However, the scarcity of such data presents a significant hurdle to extending these methods to general use cases. To address this concern, we present an automated framework to decompose trajectory data into temporally bounded and natural language-based descriptive sub-tasks by leveraging recent prompting strategies for Foundation Models (FMs) including both Large Language Models (LLMs) and Vision Language Models (VLMs). Our framework provides both time-based and language-based descriptions for lower-level sub-tasks that comprise full trajectories. To rigorously evaluate the quality of our automatic labeling framework, we contribute an algorithm SIMILARITY to produce two novel metrics, temporal similarity and semantic similarity. The metrics measure the temporal alignment and semantic fidelity of language descriptions between two sub-task decompositions, namely an FM sub-task decomposition prediction and a ground-truth sub-task decomposition. We present scores for temporal similarity and semantic similarity above 90%, compared to 30% of a randomized baseline, for multiple robotic environments, demonstrating the effectiveness of our proposed framework. Our results enable building diverse, large-scale, language-supervised datasets for improved robotic TAMP.
This study explores the ability of Image Captioning (IC) models to decode masked visual content sourced from diverse datasets. Our findings reveal the IC model's capability to generate captions from masked images, closely resembling the original content. Notably, even in the presence of masks, the model adeptly crafts descriptive textual information that goes beyond what is observable in the original image-generated captions. While the decoding performance of the IC model experiences a decline with an increase in the masked region's area, the model still performs well when important regions of the image are not masked at high coverage.
Deep learning, despite its remarkable achievements, is still a young field. Like the early stages of many scientific disciplines, it is marked by the discovery of new phenomena, ad-hoc design decisions, and the lack of a uniform and compositional mathematical foundation. From the intricacies of the implementation of backpropagation, through a growing zoo of neural network architectures, to the new and poorly understood phenomena such as double descent, scaling laws or in-context learning, there are few unifying principles in deep learning. This thesis develops a novel mathematical foundation for deep learning based on the language of category theory. We develop a new framework that is a) end-to-end, b) unform, and c) not merely descriptive, but prescriptive, meaning it is amenable to direct implementation in programming languages with sufficient features. We also systematise many existing approaches, placing many existing constructions and concepts from the literature under the same umbrella. In Part I we identify and model two main properties of deep learning systems parametricity and bidirectionality by we expand on the previously defined construction of actegories and Para to study the former, and define weighted optics to study the latter. Combining them yields parametric weighted optics, a categorical model of artificial neural networks, and more. Part II justifies the abstractions from Part I, applying them to model backpropagation, architectures, and supervised learning. We provide a lens-theoretic axiomatisation of differentiation, covering not just smooth spaces, but discrete settings of boolean circuits as well. We survey existing, and develop new categorical models of neural network architectures. We formalise the notion of optimisers and lastly, combine all the existing concepts together, providing a uniform and compositional framework for supervised learning.
We present T-Rex2, a highly practical model for open-set object detection. Previous open-set object detection methods relying on text prompts effectively encapsulate the abstract concept of common objects, but struggle with rare or complex object representation due to data scarcity and descriptive limitations. Conversely, visual prompts excel in depicting novel objects through concrete visual examples, but fall short in conveying the abstract concept of objects as effectively as text prompts. Recognizing the complementary strengths and weaknesses of both text and visual prompts, we introduce T-Rex2 that synergizes both prompts within a single model through contrastive learning. T-Rex2 accepts inputs in diverse formats, including text prompts, visual prompts, and the combination of both, so that it can handle different scenarios by switching between the two prompt modalities. Comprehensive experiments demonstrate that T-Rex2 exhibits remarkable zero-shot object detection capabilities across a wide spectrum of scenarios. We show that text prompts and visual prompts can benefit from each other within the synergy, which is essential to cover massive and complicated real-world scenarios and pave the way towards generic object detection. Model API is now available at \url{https://github.com/IDEA-Research/T-Rex}.
In this paper, we study the problem of Generalized Category Discovery (GCD), which aims to cluster unlabeled data from both known and unknown categories using the knowledge of labeled data from known categories. Current GCD methods rely on only visual cues, which however neglect the multi-modality perceptive nature of human cognitive processes in discovering novel visual categories. To address this, we propose a two-phase TextGCD framework to accomplish multi-modality GCD by exploiting powerful Visual-Language Models. TextGCD mainly includes a retrieval-based text generation (RTG) phase and a cross-modality co-teaching (CCT) phase. First, RTG constructs a visual lexicon using category tags from diverse datasets and attributes from Large Language Models, generating descriptive texts for images in a retrieval manner. Second, CCT leverages disparities between textual and visual modalities to foster mutual learning, thereby enhancing visual GCD. In addition, we design an adaptive class aligning strategy to ensure the alignment of category perceptions between modalities as well as a soft-voting mechanism to integrate multi-modality cues. Experiments on eight datasets show the large superiority of our approach over state-of-the-art methods. Notably, our approach outperforms the best competitor, by 7.7% and 10.8% in All accuracy on ImageNet-1k and CUB, respectively.
Acquiring the desired font for various design tasks can be challenging and requires professional typographic knowledge. While previous font retrieval or generation works have alleviated some of these difficulties, they often lack support for multiple languages and semantic attributes beyond the training data domains. To solve this problem, we present FontCLIP: a model that connects the semantic understanding of a large vision-language model with typographical knowledge. We integrate typography-specific knowledge into the comprehensive vision-language knowledge of a pretrained CLIP model through a novel finetuning approach. We propose to use a compound descriptive prompt that encapsulates adaptively sampled attributes from a font attribute dataset focusing on Roman alphabet characters. FontCLIP's semantic typographic latent space demonstrates two unprecedented generalization abilities. First, FontCLIP generalizes to different languages including Chinese, Japanese, and Korean (CJK), capturing the typographical features of fonts across different languages, even though it was only finetuned using fonts of Roman characters. Second, FontCLIP can recognize the semantic attributes that are not presented in the training data. FontCLIP's dual-modality and generalization abilities enable multilingual and cross-lingual font retrieval and letter shape optimization, reducing the burden of obtaining desired fonts.
Electrocardiograms (ECGs) are non-invasive diagnostic tools crucial for detecting cardiac arrhythmic diseases in clinical practice. While ECG Self-supervised Learning (eSSL) methods show promise in representation learning from unannotated ECG data, they often overlook the clinical knowledge that can be found in reports. This oversight and the requirement for annotated samples for downstream tasks limit eSSL's versatility. In this work, we address these issues with the Multimodal ECG Representation Learning (MERL}) framework. Through multimodal learning on ECG records and associated reports, MERL is capable of performing zero-shot ECG classification with text prompts, eliminating the need for training data in downstream tasks. At test time, we propose the Clinical Knowledge Enhanced Prompt Engineering (CKEPE) approach, which uses Large Language Models (LLMs) to exploit external expert-verified clinical knowledge databases, generating more descriptive prompts and reducing hallucinations in LLM-generated content to boost zero-shot classification. Based on MERL, we perform the first benchmark across six public ECG datasets, showing the superior performance of MERL compared against eSSL methods. Notably, MERL achieves an average AUC score of 75.2% in zero-shot classification (without training data), 3.2% higher than linear probed eSSL methods with 10\% annotated training data, averaged across all six datasets. Code and models are available at https://github.com/cheliu-computation/MERL