Large-scale diffusion neural networks represent a substantial milestone in text-to-image generation, but they remain poorly understood, lacking interpretability analyses. In this paper, we perform a text-image attribution analysis on Stable Diffusion, a recently open-sourced model. To produce pixel-level attribution maps, we upscale and aggregate cross-attention word-pixel scores in the denoising subnetwork, naming our method DAAM. We evaluate its correctness by testing its semantic segmentation ability on nouns, as well as its generalized attribution quality on all parts of speech, rated by humans. We then apply DAAM to study the role of syntax in the pixel space, characterizing head--dependent heat map interaction patterns for ten common dependency relations. Finally, we study several semantic phenomena using DAAM, with a focus on feature entanglement, where we find that cohyponyms worsen generation quality and descriptive adjectives attend too broadly. To our knowledge, we are the first to interpret large diffusion models from a visuolinguistic perspective, which enables future lines of research. Our code is at https://github.com/castorini/daam.
Understanding human tasks through video observations is an essential capability of intelligent agents. The challenges of such capability lie in the difficulty of generating a detailed understanding of situated actions, their effects on object states (i.e., state changes), and their causal dependencies. These challenges are further aggravated by the natural parallelism from multi-tasking and partial observations in multi-agent collaboration. Most prior works leverage action localization or future prediction as an indirect metric for evaluating such task understanding from videos. To make a direct evaluation, we introduce the EgoTaskQA benchmark that provides a single home for the crucial dimensions of task understanding through question-answering on real-world egocentric videos. We meticulously design questions that target the understanding of (1) action dependencies and effects, (2) intents and goals, and (3) agents' beliefs about others. These questions are divided into four types, including descriptive (what status?), predictive (what will?), explanatory (what caused?), and counterfactual (what if?) to provide diagnostic analyses on spatial, temporal, and causal understandings of goal-oriented tasks. We evaluate state-of-the-art video reasoning models on our benchmark and show their significant gaps between humans in understanding complex goal-oriented egocentric videos. We hope this effort will drive the vision community to move onward with goal-oriented video understanding and reasoning.
Approximate computing (AxC) has been long accepted as a design alternative for efficient system implementation at the cost of relaxed accuracy requirements. Despite the AxC research activities in various application domains, AxC thrived the past decade when it was applied in Machine Learning (ML). The by definition approximate notion of ML models but also the increased computational overheads associated with ML applications-that were effectively mitigated by corresponding approximations-led to a perfect matching and a fruitful synergy. AxC for AI/ML has transcended beyond academic prototypes. In this work, we enlighten the synergistic nature of AxC and ML and elucidate the impact of AxC in designing efficient ML systems. To that end, we present an overview and taxonomy of AxC for ML and use two descriptive application scenarios to demonstrate how AxC boosts the efficiency of ML systems.
Scene Graphs are widely applied in computer vision as a graphical representation of relationships between objects shown in images. However, these applications have not yet reached a practical stage of development owing to biased training caused by long-tailed predicate distributions. In recent years, many studies have tackled this problem. In contrast, relatively few works have considered predicate similarities as a unique dataset feature which also leads to the biased prediction. Due to the feature, infrequent predicates (e.g., parked on, covered in) are easily misclassified as closely-related frequent predicates (e.g., on, in). Utilizing predicate similarities, we propose a new classification scheme that branches the process to several fine-grained classifiers for similar predicate groups. The classifiers aim to capture the differences among similar predicates in detail. We also introduce the idea of transfer learning to enhance the features for the predicates which lack sufficient training samples to learn the descriptive representations. The results of extensive experiments on the Visual Genome dataset show that the combination of our method and an existing debiasing approach greatly improves performance on tail predicates in challenging SGCls/SGDet tasks. Nonetheless, the overall performance of the proposed approach does not reach that of the current state of the art, so further analysis remains necessary as future work.
In this paper, we consider a novel research problem: music-to-text synaesthesia. Different from the classical music tagging problem that classifies a music recording into pre-defined categories, music-to-text synaesthesia aims to generate descriptive texts from music recordings with the same sentiment for further understanding. As existing music-related datasets do not contain the semantic descriptions on music recordings, we collect a new dataset that contains 1,955 aligned pairs of classical music recordings and text descriptions. Based on this, we build a computational model to generate sentences that can describe the content of the music recording. To tackle the highly non-discriminative classical music, we design a group topology-preservation loss, which considers more samples as a group reference and preserves the relative topology among different samples. Extensive experimental results qualitatively and quantitatively demonstrate the effectiveness of our proposed model over five heuristics or pre-trained competitive methods and their variants on our collected dataset.
We tackle the problem of generating audio samples conditioned on descriptive text captions. In this work, we propose AaudioGen, an auto-regressive generative model that generates audio samples conditioned on text inputs. AudioGen operates on a learnt discrete audio representation. The task of text-to-audio generation poses multiple challenges. Due to the way audio travels through a medium, differentiating ``objects'' can be a difficult task (e.g., separating multiple people simultaneously speaking). This is further complicated by real-world recording conditions (e.g., background noise, reverberation, etc.). Scarce text annotations impose another constraint, limiting the ability to scale models. Finally, modeling high-fidelity audio requires encoding audio at high sampling rate, leading to extremely long sequences. To alleviate the aforementioned challenges we propose an augmentation technique that mixes different audio samples, driving the model to internally learn to separate multiple sources. We curated 10 datasets containing different types of audio and text annotations to handle the scarcity of text-audio data points. For faster inference, we explore the use of multi-stream modeling, allowing the use of shorter sequences while maintaining a similar bitrate and perceptual quality. We apply classifier-free guidance to improve adherence to text. Comparing to the evaluated baselines, AudioGen outperforms over both objective and subjective metrics. Finally, we explore the ability of the proposed method to generate audio continuation conditionally and unconditionally. Samples: https://felixkreuk.github.io/audiogen
Pre-training on large-scale video data has become a common recipe for learning transferable spatiotemporal representations in recent years. Despite some progress, existing methods are mostly limited to highly curated datasets (e.g., K400) and exhibit unsatisfactory out-of-the-box representations. We argue that it is due to the fact that they only capture pixel-level knowledge rather than spatiotemporal semantics, which hinders further progress in video understanding. Inspired by the great success of image-text pre-training (e.g., CLIP), we take the first step to exploit language semantics to boost transferable spatiotemporal representation learning. We introduce a new pretext task, Turning to Video for Transcript Sorting (TVTS), which sorts shuffled ASR scripts by attending to learned video representations. We do not rely on descriptive captions and learn purely from video, i.e., leveraging the natural transcribed speech knowledge to provide noisy but useful semantics over time. Our method enforces the vision model to contextualize what is happening over time so that it can re-organize the narrative transcripts, and can seamlessly apply to large-scale uncurated video data in the real world. Our method demonstrates strong out-of-the-box spatiotemporal representations on diverse benchmarks, e.g., +13.6% gains over VideoMAE on SSV2 via linear probing. The code is available at https://github.com/TencentARC/TVTS.
This research is intended to accomplish two goals: The first goal is to curate a large and information rich dataset that contains crucial and succinct summaries on the players' actions and positions and the back-and-forth travel patterns of the volleyball in professional and NCAA Div-I indoor volleyball games. While several prior studies have aimed to create similar datasets for other sports (e.g. badminton and soccer), creating such a dataset for indoor volleyball is not yet realized. The second goal is to introduce a volleyball descriptive language to fully describe the rally processes in the games and apply the language to our dataset. Based on the curated dataset and our descriptive sports language, we introduce three tasks for automated volleyball action and tactic analysis using our dataset: (1) Volleyball Rally Prediction, aimed at predicting the outcome of a rally and helping players and coaches improve decision-making in practice, (2) Setting Type and Hitting Type Prediction, to help coaches and players prepare more effectively for the game, and (3) Volleyball Tactics and Attacking Zone Statistics, to provide advanced volleyball statistics and help coaches understand the game and opponent's tactics better. We conducted case studies to show how experimental results can provide insights to the volleyball analysis community. Furthermore, experimental evaluation based on real-world data establishes a baseline for future studies and applications of our dataset and language. This study bridges the gap between the indoor volleyball field and computer science. The dataset is available at: https://github.com/haotianxia/VREN.
The automatic clinical caption generation problem is referred to as proposed model combining the analysis of frontal chest X-Ray scans with structured patient information from the radiology records. We combine two language models, the Show-Attend-Tell and the GPT-3, to generate comprehensive and descriptive radiology records. The proposed combination of these models generates a textual summary with the essential information about pathologies found, their location, and the 2D heatmaps localizing each pathology on the original X-Ray scans. The proposed model is tested on two medical datasets, the Open-I, MIMIC-CXR, and the general-purpose MS-COCO. The results measured with the natural language assessment metrics prove their efficient applicability to the chest X-Ray image captioning.
Writing reports by analyzing medical images is error-prone for inexperienced practitioners and time consuming for experienced ones. In this work, we present RepsNet that adapts pre-trained vision and language models to interpret medical images and generate automated reports in natural language. RepsNet consists of an encoder-decoder model: the encoder aligns the images with natural language descriptions via contrastive learning, while the decoder predicts answers by conditioning on encoded images and prior context of descriptions retrieved by nearest neighbor search. We formulate the problem in a visual question answering setting to handle both categorical and descriptive natural language answers. We perform experiments on two challenging tasks of medical visual question answering (VQA-Rad) and report generation (IU-Xray) on radiology image datasets. Results show that RepsNet outperforms state-of-the-art methods with 81.08 % classification accuracy on VQA-Rad 2018 and 0.58 BLEU-1 score on IU-Xray. Supplementary details are available at https://sites.google.com/view/repsnet
In this paper we describe the Portuguese-language podcast dataset we have released for academic research purposes. We give an overview of how the data was sampled, descriptive statistics over the collection, as well as information about the distribution over Brazilian and Portuguese dialects. We give results from experiments on multi-lingual summarization, showing that summarizing podcast transcripts can be performed well by a system supporting both English and Portuguese. We also show experiments on Portuguese podcast genre classification using text metadata. Combining this collection with previously released English-language collection opens up the potential for multi-modal, multi-lingual and multi-dialect podcast information access research.
The success of Zero-shot Action Recognition (ZSAR) methods is intrinsically related to the nature of semantic side information used to transfer knowledge, although this aspect has not been primarily investigated in the literature. This work introduces a new ZSAR method based on the relationships of actions-objects and actions-descriptive sentences. We demonstrate that representing all object classes using descriptive sentences generates an accurate object-action affinity estimation when a paraphrase estimation method is used as an embedder. We also show how to estimate probabilities over the set of action classes based only on a set of sentences without hard human labeling. In our method, the probabilities from these two global classifiers (i.e., which use features computed over the entire video) are combined, producing an efficient transfer knowledge model for action classification. Our results are state-of-the-art in the Kinetics-400 dataset and are competitive on UCF-101 under the ZSAR evaluation. Our code is available at https://github.com/valterlej/objsentzsar
Multiple business scenarios require an automated generation of descriptive human-readable text from structured input data. Hence, fact-to-text generation systems have been developed for various downstream tasks like generating soccer reports, weather and financial reports, medical reports, person biographies, etc. Unfortunately, previous work on fact-to-text (F2T) generation has focused primarily on English mainly due to the high availability of relevant datasets. Only recently, the problem of cross-lingual fact-to-text (XF2T) was proposed for generation across multiple languages alongwith a dataset, XALIGN for eight languages. However, there has been no rigorous work on the actual XF2T generation problem. We extend XALIGN dataset with annotated data for four more languages: Punjabi, Malayalam, Assamese and Oriya. We conduct an extensive study using popular Transformer-based text generation models on our extended multi-lingual dataset, which we call XALIGNV2. Further, we investigate the performance of different text generation strategies: multiple variations of pretraining, fact-aware embeddings and structure-aware input encoding. Our extensive experiments show that a multi-lingual mT5 model which uses fact-aware embeddings with structure-aware input encoding leads to best results on average across the twelve languages. We make our code, dataset and model publicly available, and hope that this will help advance further research in this critical area.
AI-generated synthetic data allows to distill the general patterns of existing data, that can then be shared safely as granular-level representative, yet novel data samples within the original semantics. In this work we explore approaches of incorporating domain expertise into the data synthesis, to have the statistical properties as well as pre-existing domain knowledge of rules be represented. The resulting synthetic data generator, that can be probed for any number of new samples, can then serve as a common source of intelligence, as a lingua franca of learning, consumable by humans and machines alike. We demonstrate the concept for a publicly available data set, and evaluate its benefits via descriptive analysis as well as a downstream ML model.
Multi-Document Scientific Summarization (MDSS) aims to produce coherent and concise summaries for clusters of topic-relevant scientific papers. This task requires precise understanding of paper content and accurate modeling of cross-paper relationships. Knowledge graphs convey compact and interpretable structured information for documents, which makes them ideal for content modeling and relationship modeling. In this paper, we present KGSum, an MDSS model centred on knowledge graphs during both the encoding and decoding process. Specifically, in the encoding process, two graph-based modules are proposed to incorporate knowledge graph information into paper encoding, while in the decoding process, we propose a two-stage decoder by first generating knowledge graph information of summary in the form of descriptive sentences, followed by generating the final summary. Empirical results show that the proposed architecture brings substantial improvements over baselines on the Multi-Xscience dataset.
We exploit a city-level panel comprised of individual house price estimates to estimate the impact of COVID-19 on both small and big real-estate markets in California USA. Descriptive analysis of spot house price estimates, including contemporaneous price uncertainty and 30-day price change for individual properties listed on the online real-estate platform Zillow.com, together facilitate quantifying both the excess valuation and valuation confidence attributable to this global socio-economic shock. Our quasi-experimental pre-/post-COVID-19 design spans several years around 2020 and leverages contemporaneous price estimates of rental properties - i.e., real estate entering the habitation market, just not for purchase (off-market) and hence free of speculation - as an appropriate counterfactual to properties listed for sale, which are subject to on-market speculation. Combining unit-level matching and multivariate difference-in-difference regression approaches, we obtain consistent estimates regarding the sign and magnitude of excess price growth observed after the pandemic onset. Specifically, our results indicate that properties listed for sale appreciated an additional 1% per month above what would be expected in the absence of the pandemic. This corresponds to an excess annual price growth of roughly 12.7 percentage points, which accounts for more than half of the actual annual price growth in 2021 observed across the studied regions. Simultaneously, uncertainty in price estimates decreased, signaling the irrational confidence characteristic of prior asset bubbles. We explore how these two trends are related to market size, local market supply and borrowing costs, which altogether lend support for the counterintuitive roles of uncertainty and interruptions in decision-making.
Open-vocabulary models are a promising new paradigm for image classification. Unlike traditional classification models, open-vocabulary models classify among any arbitrary set of categories specified with natural language during inference. This natural language, called "prompts", typically consists of a set of hand-written templates (e.g., "a photo of a {}") which are completed with each of the category names. This work introduces a simple method to generate higher accuracy prompts, without relying on any explicit knowledge of the task domain and with far fewer hand-constructed sentences. To achieve this, we combine open-vocabulary models with large language models (LLMs) to create Customized Prompts via Language models (CuPL, pronounced "couple"). In particular, we leverage the knowledge contained in LLMs in order to generate many descriptive sentences that contain important discriminating characteristics of the image categories. This allows the model to place a greater importance on these regions in the image when making predictions. We find that this straightforward and general approach improves accuracy on a range of zero-shot image classification benchmarks, including over one percentage point gain on ImageNet. Finally, this simple baseline requires no additional training and remains completely zero-shot. Code available at https://github.com/sarahpratt/CuPL.
This article provides a background and descriptive analysis of insect memory and the coding of olfactory sensation in Drosophila, presenting graphs and summary statistics from a large dataset of neurons and synapses that was recently made publicly available and also discussing findings from the existing empirical literature. Some general principles from Drosophila olfaction are discussed as they apply to the design of analogous systems in artificial neural networks: (1) the networks used for coding are shallow; (2) the level of connectedness varies widely across neurons in the same layer; (3) much communication is between neurons in the same layer; (4) in most olfactory learning, the manner in which sensory inputs are represented in stored memory is largely fixed, and the learning process involves developing positive or negative associations with existing categories of inputs.
Customer churn prediction is a valuable task in many industries. In telecommunications it presents great challenges, given the high dimensionality of the data, and how difficult it is to identify underlying frustration signatures, which may represent an important driver regarding future churn behaviour. Here, we propose a novel Bayesian hierarchical joint model that is able to characterise customer profiles based on how many events take place within different television watching journeys, and how long it takes between events. The model drastically reduces the dimensionality of the data from thousands of observations per customer to 11 customer-level parameter estimates and random effects. We test our methodology using data from 40 BT customers (20 active and 20 who eventually cancelled their subscription) whose TV watching behaviours were recorded from October to December 2019, totalling approximately half a million observations. Employing different machine learning techniques using the parameter estimates and random effects from the Bayesian hierarchical model as features yielded up to 92\% accuracy predicting churn, associated with 100\% true positive rates and false positive rates as low as 14\% on a validation set. Our proposed methodology represents an efficient way of reducing the dimensionality of the data, while at the same time maintaining high descriptive and predictive capabilities. We provide code to implement the Bayesian model at https://github.com/rafamoral/profiling_tv_watching_behaviour.
Today, data collection has improved in various areas, and the medical domain is no exception. Auscultation, as an important diagnostic technique for physicians, due to the progress and availability of digital stethoscopes, lends itself well to applications of machine learning. Due to the large number of auscultations performed, the availability of data opens up an opportunity for more effective analysis of sounds where prognostic accuracy even among experts remains low. In this study, digital 6-channel auscultations of 45 patients were used in various machine learning scenarios, with the aim of distinguishing between normal and anomalous pulmonary sounds. Audio features (such as fundamental frequencies F0-4, loudness, HNR, DFA, as well as descriptive statistics of log energy, RMS and MFCC) were extracted using the Python library Surfboard. Windowing and feature aggregation and concatenation strategies were used to prepare data for tree-based ensemble models in unsupervised (fair-cut forest) and supervised (random forest) machine learning settings. The evaluation was carried out using 9-fold stratified cross-validation repeated 30 times. Decision fusion by averaging outputs for a subject was tested and found to be useful. Supervised models showed a consistent advantage over unsupervised ones, achieving mean AUC ROC of 0.691 (accuracy 71.11%, Kappa 0.416, F1-score 0.771) in side-based detection and mean AUC ROC of 0.721 (accuracy 68.89%, Kappa 0.371, F1-score 0.650) in patient-based detection.