CLIP (Contrastive Language-Image Pre-training) is a very recent multi-modal model that jointly learns representations of images and texts. The model is trained on a massive amount of English data and shows impressive performance on zero-shot classification tasks. Training the same model on a different language is not trivial, since data in other languages might be not enough and the model needs high-quality translations of the texts to guarantee a good performance. In this paper, we present the first CLIP model for the Italian Language (CLIP-Italian), trained on more than 1.4 million image-text pairs. Results show that CLIP-Italian outperforms the multilingual CLIP model on the tasks of image retrieval and zero-shot classification.
Accurate visual re-localization is very critical to many artificial intelligence applications, such as augmented reality, virtual reality, robotics and autonomous driving. To accomplish this task, we propose an integrated visual re-localization method called RLOCS by combining image retrieval, semantic consistency and geometry verification to achieve accurate estimations. The localization pipeline is designed as a coarse-to-fine paradigm. In the retrieval part, we cascade the architecture of ResNet101-GeM-ArcFace and employ DBSCAN followed by spatial verification to obtain a better initial coarse pose. We design a module called observation constraints, which combines geometry information and semantic consistency for filtering outliers. Comprehensive experiments are conducted on open datasets, including retrieval on R-Oxford5k and R-Paris6k, semantic segmentation on Cityscapes, localization on Aachen Day-Night and InLoc. By creatively modifying separate modules in the total pipeline, our method achieves many performance improvements on the challenging localization benchmarks.
To build a robust and practical content-based image retrieval (CBIR) system that is applicable to a clinical brain MRI database, we propose a new framework -- Disease-oriented image embedding with pseudo-scanner standardization (DI-PSS) -- that consists of two core techniques, data harmonization and a dimension reduction algorithm. Our DI-PSS uses skull stripping and CycleGAN-based image transformations that map to a standard brain followed by transformation into a brain image taken with a given reference scanner. Then, our 3D convolutioinal autoencoders (3D-CAE) with deep metric learning acquires a low-dimensional embedding that better reflects the characteristics of the disease. The effectiveness of our proposed framework was tested on the T1-weighted MRIs selected from the Alzheimer's Disease Neuroimaging Initiative and the Parkinson's Progression Markers Initiative. We confirmed that our PSS greatly reduced the variability of low-dimensional embeddings caused by different scanner and datasets. Compared with the baseline condition, our PSS reduced the variability in the distance from Alzheimer's disease (AD) to clinically normal (CN) and Parkinson disease (PD) cases by 15.8-22.6% and 18.0-29.9%, respectively. These properties allow DI-PSS to generate lower dimensional representations that are more amenable to disease classification. In AD and CN classification experiments based on spectral clustering, PSS improved the average accuracy and macro-F1 by 6.2% and 10.7%, respectively. Given the potential of the DI-PSS for harmonizing images scanned by MRI scanners that were not used to scan the training data, we expect that the DI-PSS is suitable for application to a large number of legacy MRIs scanned in heterogeneous environments.
Hashing technology has been widely used in image retrieval due to its computational and storage efficiency. Recently, deep unsupervised hashing methods have attracted increasing attention due to the high cost of human annotations in the real world and the superiority of deep learning technology. However, most deep unsupervised hashing methods usually pre-compute a similarity matrix to model the pairwise relationship in the pre-trained feature space. Then this similarity matrix would be used to guide hash learning, in which most of the data pairs are treated equivalently. The above process is confronted with the following defects: 1) The pre-computed similarity matrix is inalterable and disconnected from the hash learning process, which cannot explore the underlying semantic information. 2) The informative data pairs may be buried by the large number of less-informative data pairs. To solve the aforementioned problems, we propose a Deep Self-Adaptive Hashing (DSAH) model to adaptively capture the semantic information with two special designs: Adaptive Neighbor Discovery (AND) and Pairwise Information Content (PIC). Firstly, we adopt the AND to initially construct a neighborhood-based similarity matrix, and then refine this initial similarity matrix with a novel update strategy to further investigate the semantic structure behind the learned representation. Secondly, we measure the priorities of data pairs with PIC and assign adaptive weights to them, which is relies on the assumption that more dissimilar data pairs contain more discriminative information for hash learning. Extensive experiments on several datasets demonstrate that the above two technologies facilitate the deep hashing model to achieve superior performance.
Deep hashing has been widely applied to large-scale image retrieval by encoding high-dimensional data points into binary codes for efficient retrieval. Compared with pairwise/triplet similarity based hash learning, central similarity based hashing can more efficiently capture the global data distribution. For multi-label image retrieval, however, previous methods only use multiple hash centers with equal weights to generate one centroid as the learning target, which ignores the relationship between the weights of hash centers and the proportion of instance regions in the image. To address the above issue, we propose a two-step alternative optimization approach, Instance-weighted Central Similarity (ICS), to automatically learn the center weight corresponding to a hash code. Firstly, we apply the maximum entropy regularizer to prevent one hash center from dominating the loss function, and compute the center weights via projection gradient descent. Secondly, we update neural network parameters by standard back-propagation with fixed center weights. More importantly, the learned center weights can well reflect the proportion of foreground instances in the image. Our method achieves the state-of-the-art performance on the image retrieval benchmarks, and especially improves the mAP by 1.6%-6.4% on the MS COCO dataset.
We explore the zero-shot setting for day-night domain adaptation. The traditional domain adaptation setting is to train on one domain and adapt to the target domain by exploiting unlabeled data samples from the test set. As gathering relevant test data is expensive and sometimes even impossible, we remove any reliance on test data imagery and instead exploit a visual inductive prior derived from physics-based reflection models for domain adaptation. We cast a number of color invariant edge detectors as trainable layers in a convolutional neural network and evaluate their robustness to illumination changes. We show that the color invariant layer reduces the day-night distribution shift in feature map activations throughout the network. We demonstrate improved performance for zero-shot day to night domain adaptation on both synthetic as well as natural datasets in various tasks, including classification, segmentation and place recognition.
Researchers often spend weeks sifting through decades of unlabeled satellite imagery(on NASA Worldview) in order to develop datasets on which they can start conducting research. We developed an interactive, scalable and fast image similarity search engine (which can take one or more images as the query image) that automatically sifts through the unlabeled dataset reducing dataset generation time from weeks to minutes. In this work, we describe key components of the end to end pipeline. Our similarity search system was created to be able to identify similar images from a potentially petabyte scale database that are similar to an input image, and for this we had to break down each query image into its features, which were generated by a classification layer stripped CNN trained in a supervised manner. To store and search these features efficiently, we had to make several scalability improvements. To improve the speed, reduce the storage, and shrink memory requirements for embedding search, we add a fully connected layer to our CNN make all images into a 128 length vector before entering the classification layers. This helped us compress the size of our image features from 2048 (for ResNet, which was initially tried as our featurizer) to 128 for our new custom model. Additionally, we utilize existing approximate nearest neighbor search libraries to significantly speed up embedding search. Our system currently searches over our entire database of images at 5 seconds per query on a single virtual machine in the cloud. In the future, we would like to incorporate a SimCLR based featurizing model which could be trained without any labelling by a human (since the classification aspect of the model is irrelevant to this use case).
We extend the task of composed image retrieval, where an input query consists of an image and short textual description of how to modify the image. Existing methods have only been applied to non-complex images within narrow domains, such as fashion products, thereby limiting the scope of study on in-depth visual reasoning in rich image and language contexts. To address this issue, we collect the Compose Image Retrieval on Real-life images (CIRR) dataset, which consists of over 36,000 pairs of crowd-sourced, open-domain images with human-generated modifying text. To extend current methods to the open-domain, we propose CIRPLANT, a transformer based model that leverages rich pre-trained vision-and-language (V&L) knowledge for modifying visual features conditioned on natural language. Retrieval is then done by nearest neighbor lookup on the modified features. We demonstrate that with a relatively simple architecture, CIRPLANT outperforms existing methods on open-domain images, while matching state-of-the-art accuracy on the existing narrow datasets, such as fashion. Together with the release of CIRR, we believe this work will inspire further research on composed image retrieval.
Image Retrieval is a fundamental task of obtaining images similar to the query one from a database. A common image retrieval practice is to firstly retrieve candidate images via similarity search using global image features and then re-rank the candidates by leveraging their local features. Previous learning-based studies mainly focus on either global or local image representation learning to tackle the retrieval task. In this paper, we abandon the two-stage paradigm and seek to design an effective single-stage solution by integrating local and global information inside images into compact image representations. Specifically, we propose a Deep Orthogonal Local and Global (DOLG) information fusion framework for end-to-end image retrieval. It attentively extracts representative local information with multi-atrous convolutions and self-attention at first. Components orthogonal to the global image representation are then extracted from the local information. At last, the orthogonal components are concatenated with the global representation as a complementary, and then aggregation is performed to generate the final representation. The whole framework is end-to-end differentiable and can be trained with image-level labels. Extensive experimental results validate the effectiveness of our solution and show that our model achieves state-of-the-art image retrieval performances on Revisited Oxford and Paris datasets.
Numerous online stock image libraries offer high quality yet copyright free images for use in marketing campaigns. To assist advertisers in navigating such third party libraries, we study the problem of automatically fetching relevant ad images given the ad text (via a short textual query for images). Motivated by our observations in logged data on ad image search queries (given ad text), we formulate a keyword extraction problem, where a keyword extracted from the ad text (or its augmented version) serves as the ad image query. In this context, we propose VisualTextRank: an unsupervised method to (i) augment input ad text using semantically similar ads, and (ii) extract the image query from the augmented ad text. VisualTextRank builds on prior work on graph based context extraction (biased TextRank in particular) by leveraging both the text and image of similar ads for better keyword extraction, and using advertiser category specific biasing with sentence-BERT embeddings. Using data collected from the Verizon Media Native (Yahoo Gemini) ad platform's stock image search feature for onboarding advertisers, we demonstrate the superiority of VisualTextRank compared to competitive keyword extraction baselines (including an $11\%$ accuracy lift over biased TextRank). For the case when the stock image library is restricted to English queries, we show the effectiveness of VisualTextRank on multilingual ads (translated to English) while leveraging semantically similar English ads. Online tests with a simplified version of VisualTextRank led to a 28.7% increase in the usage of stock image search, and a 41.6% increase in the advertiser onboarding rate in the Verizon Media Native ad platform.
Online hashing methods usually learn the hash functions online, aiming to efficiently adapt to the data variations in the streaming environment. However, when the hash functions are updated, the binary codes for the whole database have to be updated to be consistent with the hash functions, resulting in the inefficiency in the online image retrieval process. In this paper, we propose a novel online hashing framework without updating binary codes. In the proposed framework, the hash functions are fixed and a parametric similarity function for the binary codes is learnt online to adapt to the streaming data. Specifically, a parametric similarity function that has a bilinear form is adopted and a metric learning algorithm is proposed to learn the similarity function online based on the characteristics of the hashing methods. The experiments on two multi-label image datasets show that our method is competitive or outperforms the state-of-the-art online hashing methods in terms of both accuracy and efficiency for multi-label image retrieval.
In this paper, we introduce iART: an open Web platform for art-historical research that facilitates the process of comparative vision. The system integrates various machine learning techniques for keyword- and content-based image retrieval as well as category formation via clustering. An intuitive GUI supports users to define queries and explore results. By using a state-of-the-art cross-modal deep learning approach, it is possible to search for concepts that were not previously detected by trained classification models. Art-historical objects from large, openly licensed collections such as Amsterdam Rijksmuseum and Wikidata are made available to users.
We present a new human-human dialogue dataset - PhotoChat, the first dataset that casts light on the photo sharing behavior in onlin emessaging. PhotoChat contains 12k dialogues, each of which is paired with a user photo that is shared during the conversation. Based on this dataset, we propose two tasks to facilitate research on image-text modeling: a photo-sharing intent prediction task that predicts whether one intends to share a photo in the next conversation turn, and a photo retrieval task that retrieves the most relevant photo according to the dialogue context. In addition, for both tasks, we provide baseline models using the state-of-the-art models and report their benchmark performances. The best image retrieval model achieves 10.4% recall@1 (out of 1000 candidates) and the best photo intent prediction model achieves 58.1% F1 score, indicating that the dataset presents interesting yet challenging real-world problems. We are releasing PhotoChat to facilitate future research work among the community.
Person re-identification (Re-ID) aims to match pedestrians under dis-joint cameras. Most Re-ID methods formulate it as visual representation learning and image search, and its accuracy is consequently affected greatly by the search space. Spatial-temporal information has been proven to be efficient to filter irrelevant negative samples and significantly improve Re-ID accuracy. However, existing spatial-temporal person Re-ID methods are still rough and do not exploit spatial-temporal information sufficiently. In this paper, we propose a novel Instance-level and Spatial-Temporal Disentangled Re-ID method (InSTD), to improve Re-ID accuracy. In our proposed framework, personalized information such as moving direction is explicitly considered to further narrow down the search space. Besides, the spatial-temporal transferring probability is disentangled from joint distribution to marginal distribution, so that outliers can also be well modeled. Abundant experimental analyses are presented, which demonstrates the superiority and provides more insights into our method. The proposed method achieves mAP of 90.8% on Market-1501 and 89.1% on DukeMTMC-reID, improving from the baseline 82.2% and 72.7%, respectively. Besides, in order to provide a better benchmark for person re-identification, we release a cleaned data list of DukeMTMC-reID with this paper: https://github.com/RenMin1991/cleaned-DukeMTMC-reID/
The fact that there exists a gap between low-level features and semantic meanings of images, called the semantic gap, is known for decades. Resolution of the semantic gap is a long standing problem. The semantic gap problem is reviewed and a survey on recent efforts in bridging the gap is made in this work. Most importantly, we claim that the semantic gap is primarily bridged through supervised learning today. Experiences are drawn from two application domains to illustrate this point: 1) object detection and 2) metric learning for content-based image retrieval (CBIR). To begin with, this paper offers a historical retrospective on supervision, makes a gradual transition to the modern data-driven methodology and introduces commonly used datasets. Then, it summarizes various supervision methods to bridge the semantic gap in the context of object detection and metric learning.
Histopathology digital scans are large-size images that contain valuable information at the pixel level. Content-based comparison of these images is a challenging task. This study proposes a content-based similarity measure for high-resolution gigapixel histopathology images. The proposed similarity measure is an expansion of cosine vector similarity to a matrix. Each image is divided into same-size patches with a meaningful amount of information (i.e., contained enough tissue). The similarity is measured by the extraction of patch-level deep embeddings of the last pooling layer of a pre-trained deep model at four different magnification levels, namely, 1x, 2.5x, 5x, and 10x magnifications. In addition, for faster measurement, embedding reduction is investigated. Finally, to assess the proposed method, an image search method is implemented. Results show that the similarity measure represents the slide labels with a maximum accuracy of 93.18\% for top-5 search at 5x magnification.
The expanding adoption of digital pathology has enabled the curation of large repositories of histology whole slide images (WSIs), which contain a wealth of information. Similar pathology image search offers the opportunity to comb through large historical repositories of gigapixel WSIs to identify cases with similar morphological features and can be particularly useful for diagnosing rare diseases, identifying similar cases for predicting prognosis, treatment outcomes, and potential clinical trial success. A critical challenge in developing a WSI search and retrieval system is scalability, which is uniquely challenging given the need to search a growing number of slides that each can consist of billions of pixels and are several gigabytes in size. Such systems are typically slow and retrieval speed often scales with the size of the repository they search through, making their clinical adoption tedious and are not feasible for repositories that are constantly growing. Here we present Fast Image Search for Histopathology (FISH), a histology image search pipeline that is infinitely scalable and achieves constant search speed that is independent of the image database size while being interpretable and without requiring detailed annotations. FISH uses self-supervised deep learning to encode meaningful representations from WSIs and a Van Emde Boas tree for fast search, followed by an uncertainty-based ranking algorithm to retrieve similar WSIs. We evaluated FISH on multiple tasks and datasets with over 22,000 patient cases spanning 56 disease subtypes. We additionally demonstrate that FISH can be used to assist with the diagnosis of rare cancer types where sufficient cases may not be available to train traditional supervised deep models. FISH is available as an easy-to-use, open-source software package (https://github.com/mahmoodlab/FISH).
Text-to-image person re-identification (ReID) aims to search for images containing a person of interest using textual descriptions. However, due to the significant modality gap and the large intra-class variance in textual descriptions, text-to-image ReID remains a challenging problem. Accordingly, in this paper, we propose a Semantically Self-Aligned Network (SSAN) to handle the above problems. First, we propose a novel method that automatically extracts semantically aligned part-level features from the two modalities. Second, we design a multi-view non-local network that captures the relationships between body parts, thereby establishing better correspondences between body parts and noun phrases. Third, we introduce a Compound Ranking (CR) loss that makes use of textual descriptions for other images of the same identity to provide extra supervision, thereby effectively reducing the intra-class variance in textual features. Finally, to expedite future research in text-to-image ReID, we build a new database named ICFG-PEDES. Extensive experiments demonstrate that SSAN outperforms state-of-the-art approaches by significant margins. Both the new ICFG-PEDES database and the SSAN code are available at https://github.com/zifyloo/SSAN.
We address representation learning for large-scale instance-level image retrieval. Apart from backbone, training pipelines and loss functions, popular approaches have focused on different spatial pooling and attention mechanisms, which are at the core of learning a powerful global image representation. There are different forms of attention according to the interaction of elements of the feature tensor (local and global) and the dimensions where it is applied (spatial and channel). Unfortunately, each study addresses only one or two forms of attention and applies it to different problems like classification, detection or retrieval. We present global-local attention module (GLAM), which is attached at the end of a backbone network and incorporates all four forms of attention: local and global, spatial and channel. We obtain a new feature tensor and, by spatial pooling, we learn a powerful embedding for image retrieval. Focusing on global descriptors, we provide empirical evidence of the interaction of all forms of attention and improve the state of the art on standard benchmarks.
Recently, it has been observed that {0,1,-1}-ternary codes which are simply generated from deep features by hard thresholding, tend to outperform {-1,1}-binary codes in image retrieval. To obtain better ternary codes, we for the first time propose to jointly learn the features with the codes by appending a smoothed function to the networks. During training, the function could evolve into a non-smoothed ternary function by a continuation method. The method circumvents the difficulty of directly training discrete functions and reduces the quantization errors of ternary codes. Experiments show that the generated codes indeed could achieve higher retrieval accuracy.