Facial analysis has emerged as a prominent area of research with diverse applications, including cosmetic surgery programs, the beauty industry, photography, and entertainment. Manipulating patient images often necessitates professional image processing software. This study contributes by providing a model that facilitates the detection of blemishes and skin lesions on facial images through a convolutional neural network and machine learning approach. The proposed method offers advantages such as simple architecture, speed and suitability for image processing while avoiding the complexities associated with traditional methods. The model comprises four main steps: area selection, scanning the chosen region, lesion diagnosis, and marking the identified lesion. Raw data for this research were collected from a reputable clinic in Tehran specializing in skincare and beauty services. The dataset includes administrative information, clinical data, and facial and profile images. A total of 2300 patient images were extracted from this raw data. A software tool was developed to crop and label lesions, with input from two treatment experts. In the lesion preparation phase, the selected area was standardized to 50 * 50 pixels. Subsequently, a convolutional neural network model was employed for lesion labeling. The classification model demonstrated high accuracy, with a measure of 0.98 for healthy skin and 0.97 for lesioned skin specificity. Internal validation involved performance indicators and cross-validation, while external validation compared the model's performance indicators with those of the transfer learning method using the Vgg16 deep network model. Compared to existing studies, the results of this research showcase the efficacy and desirability of the proposed model and methodology.
Advancements in generative models have sparked significant interest in generating images while adhering to specific structural guidelines. Scene graph to image generation is one such task of generating images which are consistent with the given scene graph. However, the complexity of visual scenes poses a challenge in accurately aligning objects based on specified relations within the scene graph. Existing methods approach this task by first predicting a scene layout and generating images from these layouts using adversarial training. In this work, we introduce a novel approach to generate images from scene graphs which eliminates the need of predicting intermediate layouts. We leverage pre-trained text-to-image diffusion models and CLIP guidance to translate graph knowledge into images. Towards this, we first pre-train our graph encoder to align graph features with CLIP features of corresponding images using a GAN based training. Further, we fuse the graph features with CLIP embedding of object labels present in the given scene graph to create a graph consistent CLIP guided conditioning signal. In the conditioning input, object embeddings provide coarse structure of the image and graph features provide structural alignment based on relationships among objects. Finally, we fine tune a pre-trained diffusion model with the graph consistent conditioning signal with reconstruction and CLIP alignment loss. Elaborate experiments reveal that our method outperforms existing methods on standard benchmarks of COCO-stuff and Visual Genome dataset.
Causal inference from observational data has recently found many applications in machine learning. While sound and complete algorithms exist to compute causal effects, many of these algorithms require explicit access to conditional likelihoods over the observational distribution, which is difficult to estimate in the high-dimensional regime, such as with images. To alleviate this issue, researchers have approached the problem by simulating causal relations with neural models and obtained impressive results. However, none of these existing approaches can be applied to generic scenarios such as causal graphs on image data with latent confounders, or obtain conditional interventional samples. In this paper, we show that any identifiable causal effect given an arbitrary causal graph can be computed through push-forward computations of conditional generative models. Based on this result, we devise a diffusion-based approach to sample from any (conditional) interventional distribution on image data. To showcase our algorithm's performance, we conduct experiments on a Colored MNIST dataset having both the treatment ($X$) and the target variables ($Y$) as images and obtain interventional samples from $P(y|do(x))$. As an application of our algorithm, we evaluate two large conditional generative models that are pre-trained on the CelebA dataset by analyzing the strength of spurious correlations and the level of disentanglement they achieve.
Malaria is a major health issue worldwide, and its diagnosis requires scalable solutions that can work effectively with low-cost microscopes (LCM). Deep learning-based methods have shown success in computer-aided diagnosis from microscopic images. However, these methods need annotated images that show cells affected by malaria parasites and their life stages. Annotating images from LCM significantly increases the burden on medical experts compared to annotating images from high-cost microscopes (HCM). For this reason, a practical solution would be trained on HCM images which should generalize well on LCM images during testing. While earlier methods adopted a multi-stage learning process, they did not offer an end-to-end approach. In this work, we present an end-to-end learning framework, named CodaMal (Contrastive Domain Adpation for Malaria). In order to bridge the gap between HCM (training) and LCM (testing), we propose a domain adaptive contrastive loss. It reduces the domain shift by promoting similarity between the representations of HCM and its corresponding LCM image, without imposing an additional annotation burden. In addition, the training objective includes object detection objectives with carefully designed augmentations, ensuring the accurate detection of malaria parasites. On the publicly available large-scale M5-dataset, our proposed method shows a significant improvement of 16% over the state-of-the-art methods in terms of the mean average precision metric (mAP), provides 21x speed up during inference, and requires only half learnable parameters than the prior methods. Our code is publicly available.
Hairstyle reflects culture and ethnicity at first glance. In the digital era, various realistic human hairstyles are also critical to high-fidelity digital human assets for beauty and inclusivity. Yet, realistic hair modeling and real-time rendering for animation is a formidable challenge due to its sheer number of strands, complicated structures of geometry, and sophisticated interaction with light. This paper presents GaussianHair, a novel explicit hair representation. It enables comprehensive modeling of hair geometry and appearance from images, fostering innovative illumination effects and dynamic animation capabilities. At the heart of GaussianHair is the novel concept of representing each hair strand as a sequence of connected cylindrical 3D Gaussian primitives. This approach not only retains the hair's geometric structure and appearance but also allows for efficient rasterization onto a 2D image plane, facilitating differentiable volumetric rendering. We further enhance this model with the "GaussianHair Scattering Model", adept at recreating the slender structure of hair strands and accurately capturing their local diffuse color in uniform lighting. Through extensive experiments, we substantiate that GaussianHair achieves breakthroughs in both geometric and appearance fidelity, transcending the limitations encountered in state-of-the-art methods for hair reconstruction. Beyond representation, GaussianHair extends to support editing, relighting, and dynamic rendering of hair, offering seamless integration with conventional CG pipeline workflows. Complementing these advancements, we have compiled an extensive dataset of real human hair, each with meticulously detailed strand geometry, to propel further research in this field.
Depth completion is a crucial task in autonomous driving, aiming to convert a sparse depth map into a dense depth prediction. Due to its potentially rich semantic information, RGB image is commonly fused to enhance the completion effect. Image-guided depth completion involves three key challenges: 1) how to effectively fuse the two modalities; 2) how to better recover depth information; and 3) how to achieve real-time prediction for practical autonomous driving. To solve the above problems, we propose a concise but effective network, named CENet, to achieve high-performance depth completion with a simple and elegant structure. Firstly, we use a fast guidance module to fuse the two sensor features, utilizing abundant auxiliary features extracted from the color space. Unlike other commonly used complicated guidance modules, our approach is intuitive and low-cost. In addition, we find and analyze the optimization inconsistency problem for observed and unobserved positions, and a decoupled depth prediction head is proposed to alleviate the issue. The proposed decoupled head can better output the depth of valid and invalid positions with very few extra inference time. Based on the simple structure of dual-encoder and single-decoder, our CENet can achieve superior balance between accuracy and efficiency. In the KITTI depth completion benchmark, our CENet attains competitive performance and inference speed compared with the state-of-the-art methods. To validate the generalization of our method, we also evaluate on indoor NYUv2 dataset, and our CENet still achieve impressive results. The code of this work will be available at https://github.com/lmomoy/CENet.
X-ray imaging can be efficiently used for high-throughput in-line inspection of industrial products. However, designing a system that satisfies industrial requirements and achieves high accuracy is a challenging problem. The effect of many system settings is application-specific and difficult to predict in advance. Consequently, the system is often configured using empirical rules and visual observations. The performance of the resulting system is characterized by extensive experimental testing. We propose to use computational methods to substitute real measurements with generated images corresponding to the same experimental settings. With this approach, it is possible to observe the influence of experimental settings on a large amount of data and to make a prediction of the system performance faster than with conventional methods. We argue that a high accuracy of the image generator may be unnecessary for an accurate performance prediction. We propose a quantitative methodology to characterize the quality of the generation model using POD curves. The proposed approach can be adapted to various applications and we demonstrate it on the poultry inspection problem. We show how a calibrated image generation model can be used to quantitatively evaluate the effect of the X-ray exposure time on the performance of the inspection system.
Estimating the 3D shape of an object using a single image is a difficult problem. Modern approaches achieve good results for general objects, based on real photographs, but worse results on less expressive representations such as historic sketches. Our automated approach generates a variety of detailed 3D representation from a single sketch, depicting a medieval statue, and can be guided by multi-modal inputs, such as text prompts. It relies solely on synthetic data for training, making it adoptable even in cases of only small numbers of training examples. Our solution allows domain experts such as a curators to interactively reconstruct potential appearances of lost artifacts.
Vision-language models (VLMs) have emerged as formidable tools, showing their strong capability in handling various open-vocabulary tasks in image recognition, text-driven visual content generation, and visual chatbots, to name a few. In recent years, considerable efforts and resources have been devoted to adaptation methods for improving downstream performance of VLMs, particularly on parameter-efficient fine-tuning methods like prompt learning. However, a crucial aspect that has been largely overlooked is the confidence calibration problem in fine-tuned VLMs, which could greatly reduce reliability when deploying such models in the real world. This paper bridges the gap by systematically investigating the confidence calibration problem in the context of prompt learning and reveals that existing calibration methods are insufficient to address the problem, especially in the open-vocabulary setting. To solve the problem, we present a simple and effective approach called Distance-Aware Calibration (DAC), which is based on scaling the temperature using as guidance the distance between predicted text labels and base classes. The experiments with 7 distinct prompt learning methods applied across 11 diverse downstream datasets demonstrate the effectiveness of DAC, which achieves high efficacy without sacrificing the inference speed.
Reconstructing and rendering 3D objects from highly sparse views is of critical importance for promoting applications of 3D vision techniques and improving user experience. However, images from sparse views only contain very limited 3D information, leading to two significant challenges: 1) Difficulty in building multi-view consistency as images for matching are too few; 2) Partially omitted or highly compressed object information as view coverage is insufficient. To tackle these challenges, we propose GaussianObject, a framework to represent and render the 3D object with Gaussian splatting, that achieves high rendering quality with only 4 input images. We first introduce techniques of visual hull and floater elimination which explicitly inject structure priors into the initial optimization process for helping build multi-view consistency, yielding a coarse 3D Gaussian representation. Then we construct a Gaussian repair model based on diffusion models to supplement the omitted object information, where Gaussians are further refined. We design a self-generating strategy to obtain image pairs for training the repair model. Our GaussianObject is evaluated on several challenging datasets, including MipNeRF360, OmniObject3D, and OpenIllumination, achieving strong reconstruction results from only 4 views and significantly outperforming previous state-of-the-art methods.