Existing industrial anomaly detection (IAD) methods predict anomaly scores for both anomaly detection and localization. However, they struggle to perform a multi-turn dialog and detailed descriptions for anomaly regions, e.g., color, shape, and categories of industrial anomalies. Recently, large multimodal (i.e., vision and language) models (LMMs) have shown eminent perception abilities on multiple vision tasks such as image captioning, visual understanding, visual reasoning, etc., making it a competitive potential choice for more comprehensible anomaly detection. However, the knowledge about anomaly detection is absent in existing general LMMs, while training a specific LMM for anomaly detection requires a tremendous amount of annotated data and massive computation resources. In this paper, we propose a novel large multi-modal model by applying vision experts for industrial anomaly detection (dubbed Myriad), which leads to definite anomaly detection and high-quality anomaly description. Specifically, we adopt MiniGPT-4 as the base LMM and design an Expert Perception module to embed the prior knowledge from vision experts as tokens which are intelligible to Large Language Models (LLMs). To compensate for the errors and confusions of vision experts, we introduce a domain adapter to bridge the visual representation gaps between generic and industrial images. Furthermore, we propose a Vision Expert Instructor, which enables the Q-Former to generate IAD domain vision-language tokens according to vision expert prior. Extensive experiments on MVTec-AD and VisA benchmarks demonstrate that our proposed method not only performs favorably against state-of-the-art methods under the 1-class and few-shot settings, but also provide definite anomaly prediction along with detailed descriptions in IAD domain.
We present a set of metrics that utilize vision priors to effectively assess the performance of saliency methods on image classification tasks. To understand behavior in deep learning models, many methods provide visual saliency maps emphasizing image regions that most contribute to a model prediction. However, there is limited work on analyzing the reliability of saliency methods in explaining model decisions. We propose the metric COnsistency-SEnsitivity (COSE) that quantifies the equivariant and invariant properties of visual model explanations using simple data augmentations. Through our metrics, we show that although saliency methods are thought to be architecture-independent, most methods could better explain transformer-based models over convolutional-based models. In addition, GradCAM was found to outperform other methods in terms of COSE but was shown to have limitations such as lack of variability for fine-grained datasets. The duality between consistency and sensitivity allow the analysis of saliency methods from different angles. Ultimately, we find that it is important to balance these two metrics for a saliency map to faithfully show model behavior.
Inter-modal image registration (IMIR) and image segmentation with abdominal Ultrasound (US) data has many important clinical applications, including image-guided surgery, automatic organ measurement and robotic navigation. However, research is severely limited by the lack of public datasets. We propose TRUSTED (the Tridimensional Renal Ultra Sound TomodEnsitometrie Dataset), comprising paired transabdominal 3DUS and CT kidney images from 48 human patients (96 kidneys), including segmentation, and anatomical landmark annotations by two experienced radiographers. Inter-rater segmentation agreement was over 94 (Dice score), and gold-standard segmentations were generated using the STAPLE algorithm. Seven anatomical landmarks were annotated, important for IMIR systems development and evaluation. To validate the dataset's utility, 5 competitive Deep Learning models for automatic kidney segmentation were benchmarked, yielding average DICE scores from 83.2% to 89.1% for CT, and 61.9% to 79.4% for US images. Three IMIR methods were benchmarked, and Coherent Point Drift performed best with an average Target Registration Error of 4.53mm. The TRUSTED dataset may be used freely researchers to develop and validate new segmentation and IMIR methods.
The emotional theory of mind problem in images is an emotion recognition task, specifically asking "How does the person in the bounding box feel?" Facial expressions, body pose, contextual information and implicit commonsense knowledge all contribute to the difficulty of the task, making this task currently one of the hardest problems in affective computing. The goal of this work is to evaluate the emotional commonsense knowledge embedded in recent large vision language models (CLIP, LLaVA) and large language models (GPT-3.5) on the Emotions in Context (EMOTIC) dataset. In order to evaluate a purely text-based language model on images, we construct "narrative captions" relevant to emotion perception, using a set of 872 physical social signal descriptions related to 26 emotional categories, along with 224 labels for emotionally salient environmental contexts, sourced from writer's guides for character expressions and settings. We evaluate the use of the resulting captions in an image-to-language-to-emotion task. Experiments using zero-shot vision-language models on EMOTIC show that combining "fast" and "slow" reasoning is a promising way forward to improve emotion recognition systems. Nevertheless, a gap remains in the zero-shot emotional theory of mind task compared to prior work trained on the EMOTIC dataset.
Building video-language foundation models is costly and difficult due to the redundant nature of video data and the lack of high-quality video-language datasets. In this paper, we propose an efficient framework to harvest video foundation models from image ones. Our method is intuitively simple by randomly dropping input video patches and masking out input text during the post-pretraining procedure. The patch dropping boosts the training efficiency significantly and text masking enforces the learning of cross-modal fusion. We conduct extensive experiments to validate the effectiveness of our method on a wide range of video-language downstream tasks including various zero-shot tasks, video question answering, and video-text retrieval. Despite its simplicity, our method achieves state-of-the-art performances, which are comparable to some heavily pretrained video foundation models. Our method is extremely efficient and can be trained in less than one day on 8 GPUs, requiring only WebVid-10M as pretraining data. We hope our method can serve as a simple yet strong counterpart for prevalent video foundation models, provide useful insights when building them, and make large pretrained models more accessible and sustainable. This is part of the InternVideo project \url{https://github.com/OpenGVLab/InternVideo}.
Learning curve extrapolation aims to predict model performance in later epochs of training, based on the performance in earlier epochs. In this work, we argue that, while the inherent uncertainty in the extrapolation of learning curves warrants a Bayesian approach, existing methods are (i) overly restrictive, and/or (ii) computationally expensive. We describe the first application of prior-data fitted neural networks (PFNs) in this context. A PFN is a transformer, pre-trained on data generated from a prior, to perform approximate Bayesian inference in a single forward pass. We propose LC-PFN, a PFN trained to extrapolate 10 million artificial right-censored learning curves generated from a parametric prior proposed in prior art using MCMC. We demonstrate that LC-PFN can approximate the posterior predictive distribution more accurately than MCMC, while being over 10 000 times faster. We also show that the same LC-PFN achieves competitive performance extrapolating a total of 20 000 real learning curves from four learning curve benchmarks (LCBench, NAS-Bench-201, Taskset, and PD1) that stem from training a wide range of model architectures (MLPs, CNNs, RNNs, and Transformers) on 53 different datasets with varying input modalities (tabular, image, text, and protein data). Finally, we investigate its potential in the context of model selection and find that a simple LC-PFN based predictive early stopping criterion obtains 2 - 6x speed-ups on 45 of these datasets, at virtually no overhead.
Existing methods attempt to improve models' generalization ability on real-world hazy images by exploring well-designed training schemes (e.g., cycleGAN, prior loss). However, most of them need very complicated training procedures to achieve satisfactory results. In this work, we present a totally novel testing pipeline called Prompt-based Test-Time Dehazing (PTTD) to help generate visually pleasing results of real-captured hazy images during the inference phase. We experimentally find that given a dehazing model trained on synthetic data, by fine-tuning the statistics (i.e., mean and standard deviation) of encoding features, PTTD is able to narrow the domain gap, boosting the performance of real image dehazing. Accordingly, we first apply a prompt generation module (PGM) to generate a visual prompt, which is the source of appropriate statistical perturbations for mean and standard deviation. And then, we employ the feature adaptation module (FAM) into the existing dehazing models for adjusting the original statistics with the guidance of the generated prompt. Note that, PTTD is model-agnostic and can be equipped with various state-of-the-art dehazing models trained on synthetic hazy-clean pairs. Extensive experimental results demonstrate that our PTTD is flexible meanwhile achieves superior performance against state-of-the-art dehazing methods in real-world scenarios.
In this report, we introduce NICE (New frontiers for zero-shot Image Captioning Evaluation) project and share the results and outcomes of 2023 challenge. This project is designed to challenge the computer vision community to develop robust image captioning models that advance the state-of-the-art both in terms of accuracy and fairness. Through the challenge, the image captioning models were tested using a new evaluation dataset that includes a large variety of visual concepts from many domains. There was no specific training data provided for the challenge, and therefore the challenge entries were required to adapt to new types of image descriptions that had not been seen during training. This report includes information on the newly proposed NICE dataset, evaluation methods, challenge results, and technical details of top-ranking entries. We expect that the outcomes of the challenge will contribute to the improvement of AI models on various vision-language tasks.
This paper addresses the challenge of generating Counterfactual Explanations (CEs), involving the identification and modification of the fewest necessary features to alter a classifier's prediction for a given image. Our proposed method, Text-to-Image Models for Counterfactual Explanations (TIME), is a black-box counterfactual technique based on distillation. Unlike previous methods, this approach requires solely the image and its prediction, omitting the need for the classifier's structure, parameters, or gradients. Before generating the counterfactuals, TIME introduces two distinct biases into Stable Diffusion in the form of textual embeddings: the context bias, associated with the image's structure, and the class bias, linked to class-specific features learned by the target classifier. After learning these biases, we find the optimal latent code applying the classifier's predicted class token and regenerate the image using the target embedding as conditioning, producing the counterfactual explanation. Extensive empirical studies validate that TIME can generate explanations of comparable effectiveness even when operating within a black-box setting.
The reconstruction kernel in computed tomography (CT) generation determines the texture of the image. Consistency in reconstruction kernels is important as the underlying CT texture can impact measurements during quantitative image analysis. Harmonization (i.e., kernel conversion) minimizes differences in measurements due to inconsistent reconstruction kernels. Existing methods investigate harmonization of CT scans in single or multiple manufacturers. However, these methods require paired scans of hard and soft reconstruction kernels that are spatially and anatomically aligned. Additionally, a large number of models need to be trained across different kernel pairs within manufacturers. In this study, we adopt an unpaired image translation approach to investigate harmonization between and across reconstruction kernels from different manufacturers by constructing a multipath cycle generative adversarial network (GAN). We use hard and soft reconstruction kernels from the Siemens and GE vendors from the National Lung Screening Trial dataset. We use 50 scans from each reconstruction kernel and train a multipath cycle GAN. To evaluate the effect of harmonization on the reconstruction kernels, we harmonize 50 scans each from Siemens hard kernel, GE soft kernel and GE hard kernel to a reference Siemens soft kernel (B30f) and evaluate percent emphysema. We fit a linear model by considering the age, smoking status, sex and vendor and perform an analysis of variance (ANOVA) on the emphysema scores. Our approach minimizes differences in emphysema measurement and highlights the impact of age, sex, smoking status and vendor on emphysema quantification.