Composed image retrieval (CIR) is the task of retrieving a target image specified by a query image and a relative text that describes a semantic modification to the query image. Existing methods in CIR struggle to accurately represent the image and the text modification, resulting in subpar performance. To address this limitation, we introduce a CIR framework, ConText-CIR, trained with a Text Concept-Consistency loss that encourages the representations of noun phrases in the text modification to better attend to the relevant parts of the query image. To support training with this loss function, we also propose a synthetic data generation pipeline that creates training data from existing CIR datasets or unlabeled images. We show that these components together enable stronger performance on CIR tasks, setting a new state-of-the-art in composed image retrieval in both the supervised and zero-shot settings on multiple benchmark datasets, including CIRR and CIRCO. Source code, model checkpoints, and our new datasets are available at https://github.com/mvrl/ConText-CIR.

Massive-scale pretraining has made vision-language models increasingly popular for image-to-image and text-to-image retrieval across a broad collection of domains. However, these models do not perform well when used for challenging retrieval tasks, such as instance retrieval in very large-scale image collections. Recent work has shown that linear transformations of VLM features trained for instance retrieval can improve performance by emphasizing subspaces that relate to the domain of interest. In this paper, we explore a more extreme version of this specialization by learning to map a given query to a query-specific feature space transformation. Because this transformation is linear, it can be applied with minimal computational cost to millions of image embeddings, making it effective for large-scale retrieval or re-ranking. Results show that this method consistently outperforms state-of-the-art alternatives, including those that require many orders of magnitude more computation at query time.

Geospatial artificial intelligence (GeoAI) and data processing techniques have significantly advanced object detection, prediction, and classification tasks. However, the availability of machine learning-ready, labeled data for specific applications such as plant disease detection remains the major challenge for the broader adoption of GeoAI. For instance, collecting temporal unmanned aerial vehicle (UAV) imagery of agricultural crops to track disease emergence and progress requires substantial human labor and resources, which is often limited to a small spatial scale. Recognizing the pivotal role of temporal data in pattern recognition, object detection, and scene reconstruction, we introduce an innovative approach to augment multispectral temporal datasets: the geospatial time machine (GTM). Our proposed methodology combines graph neural network (GNN) and generative adversarial network (GAN) architectures to generate comprehensive synthetic temporal data encompassing multivariate time series. The results demonstrate that imagery generated through backcasting can enhance the accuracy of downstream classification tasks by up to 53% in plant disease detection, particularly in the initial stages of analyzing a crop growth using multispectral and multitemporal datasets.

Large scale field-phenotyping approaches have the potential to solve important questions about the relationship of plant genotype to plant phenotype. Computational approaches to measuring the phenotype (the observable plant features) are required to address the problem at a large scale, but machine learning approaches to extract phenotypes from sensor data have been hampered by limited access to (a) sufficiently large, organized multi-sensor datasets, (b) field trials that have a large scale and significant number of genotypes, (c) full genetic sequencing of those phenotypes, and (d) datasets sufficiently organized so that algorithm centered researchers can directly address the real biological problems. To address this, we present SGxP, a novel benchmark dataset from a large-scale field trial consisting of the complete genotype of over 300 sorghum varieties, and time sequences of imagery from several field plots growing each variety, taken with RGB and laser 3D scanner imaging. To lower the barrier to entry and facilitate further developments, we provide a set of well organized, multi-sensor imagery and corresponding genomic data. We implement baseline deep learning based phenotyping approaches to create baseline results for individual sensors and multi-sensor fusion for detecting genetic mutations with known impacts. We also provide and support an open-ended challenge by identifying thousands of genetic mutations whose phenotypic impacts are currently unknown. A web interface for machine learning researchers and practitioners to share approaches, visualizations and hypotheses supports engagement with plant biologists to further the understanding of the sorghum genotype x phenotype relationship. The full dataset, leaderboard (including baseline results) and discussion forums can be found at http://sorghumsnpbenchmark.com.

Transformer architectures have shown promise for a wide range of computer vision tasks, including image embedding. As was the case with convolutional neural networks and other models, explainability of the predictions is a key concern, but visualization approaches tend to be architecture-specific. In this paper, we introduce a new method for producing interpretable visualizations that, given a pair of images encoded with a Transformer, show which regions contributed to their similarity. Additionally, for the task of image retrieval, we compare the performance of Transformer and ResNet models of similar capacity and show that while they have similar performance in aggregate, the retrieved results and the visual explanations for those results are quite different. Code is available at https://github.com/vidarlab/xformer-paired-viz.

Near-earth hyperspectral big data present both huge opportunities and challenges for spurring developments in agriculture and high-throughput plant phenotyping and breeding. In this article, we present data-driven approaches to address the calibration challenges for utilizing near-earth hyperspectral data for agriculture. A data-driven, fully automated calibration workflow that includes a suite of robust algorithms for radiometric calibration, bidirectional reflectance distribution function (BRDF) correction and reflectance normalization, soil and shadow masking, and image quality assessments was developed. An empirical method that utilizes predetermined models between camera photon counts (digital numbers) and downwelling irradiance measurements for each spectral band was established to perform radiometric calibration. A kernel-driven semiempirical BRDF correction method based on the Ross Thick-Li Sparse (RTLS) model was used to normalize the data for both changes in solar elevation and sensor view angle differences attributed to pixel location within the field of view. Following rigorous radiometric and BRDF corrections, novel rule-based methods were developed to conduct automatic soil removal; and a newly proposed approach was used for image quality assessment; additionally, shadow masking and plot-level feature extraction were carried out. Our results show that the automated calibration, processing, storage, and analysis pipeline developed in this work can effectively handle massive amounts of hyperspectral data and address the urgent challenges related to the production of sustainable bioenergy and food crops, targeting methods to accelerate plant breeding for improving yield and biomass traits.

Visualization tools like t-SNE and UMAP give insight into the high-dimensional structure of datasets. When there are related datasets (such as the high-dimensional representations of image data created by two different Deep Learning architectures), roughly aligning those visualizations helps to highlight both the similarities and differences. In this paper we propose a method to align multiple low dimensional UMAP visualizations by adding an alignment term to the UMAP loss function. We provide an automated procedure to find a weight for this term that encourages the alignment but only minimally changes the fidelity of the underlying embedding.

Deep metric learning seeks to define an embedding where semantically similar images are embedded to nearby locations, and semantically dissimilar images are embedded to distant locations. Substantial work has focused on loss functions and strategies to learn these embeddings by pushing images from the same class as close together in the embedding space as possible. In this paper, we propose an alternative, loosened embedding strategy that requires the embedding function only map each training image to the most similar examples from the same class, an approach we call "Easy Positive" mining. We provide a collection of experiments and visualizations that highlight that this Easy Positive mining leads to embed-dings that are more flexible and generalize better to new unseen data. This simple mining strategy yields recall performance that exceeds state of the art approaches (including those with complicated loss functions and ensemble methods) on image retrieval datasets including CUB, Stanford Online Products, In-Shop Clothes and Hotels-50K. Code is available at: https://github.com/littleredxh/EasyPositiveHardNegative.

Recognizing a hotel from an image of a hotel room is important for human trafficking investigations. Images directly link victims to places and can help verify where victims have been trafficked, and where their traffickers might move them or others in the future. Recognizing the hotel from images is challenging because of low image quality, uncommon camera perspectives, large occlusions (often the victim), and the similarity of objects (e.g., furniture, art, bedding) across different hotel rooms. To support efforts towards this hotel recognition task, we have curated a dataset of over 1 million annotated hotel room images from 50,000 hotels. These images include professionally captured photographs from travel websites and crowd-sourced images from a mobile application, which are more similar to the types of images analyzed in real-world investigations. We present a baseline approach based on a standard network architecture and a collection of data-augmentation approaches tuned to this problem domain.

For convolutional neural network models that optimize an image embedding, we propose a method to highlight the regions of images that contribute most to pairwise similarity. This work is a corollary to the visualization tools developed for classification networks, but applicable to the problem domains better suited to similarity learning. The visualization shows how similarity networks that are fine-tuned learn to focus on different features. We also generalize our approach to embedding networks that use different pooling strategies and provide a simple mechanism to support image similarity searches on objects or sub-regions in the query image.

Many computer vision applications rely on matching features of a query image to reference data sets, but little work has explored how quickly data sets become out of date. In this paper we measure feature matching performance across 5 years of time-lapse data from 20 static cameras to empirically study how feature matching is affected by changing sunlight direction, seasons, weather, and the structural changes over time in outdoor settings. We identify several trends that may be relevant in real world applications: (1) features are much more likely to match within a few days of the reference data, (2) weather and sun-direction have a large effect on feature matching, and (3) there is a slow decay over time due to physical changes in a scene, but this decay is much smaller than effects of lighting direction and weather. These trends are consistent across standard choices for feature detection (DoG, MSER) and feature description (SIFT, SURF, and DAISY). Across all choices, analysis of the feature detection and matching pipeline highlights that performance decay is mostly due to failures in key point detection rather than feature description.