Collecting high quality data for object detection tasks is challenging due to the inherent subjectivity in labeling the boundaries of an object. This makes it difficult to not only collect consistent annotations across a dataset but also to validate them, as no two annotators are likely to label the same object using the exact same coordinates. These challenges are further compounded when object boundaries are partially visible or blurred, which can be the case in many domains. Training on noisy annotations significantly degrades detector performance, rendering them unusable, particularly in few-shot settings, where just a few corrupted annotations can impact model performance. In this work, we propose FMG-Det, a simple, efficient methodology for training models with noisy annotations. More specifically, we propose combining a multiple instance learning (MIL) framework with a pre-processing pipeline that leverages powerful foundation models to correct labels prior to training. This pre-processing pipeline, along with slight modifications to the detector head, results in state-of-the-art performance across a number of datasets, for both standard and few-shot scenarios, while being much simpler and more efficient than other approaches.

We present a framework for inferring an atmospheric transmission profile from a spectral scene. This framework leverages a lightweight, physics-based simulator that is automatically tuned – by virtue of autodifferentiation and differentiable programming – to construct a surrogate atmospheric profile to model the observed data. We demonstrate utility of the methodology by (i) performing atmospheric correction, (ii) recasting spectral data between various modalities (e.g. radiance and reflectance at the surface and at the sensor), and (iii) inferring atmospheric transmission profiles, such as absorbing bands and their relative magnitudes.

It is often said that a deep learning model is ``invariant'' to some specific type of transformation. However, what is meant by this statement strongly depends on the context in which it is made. In this paper we explore the nature of invariance and equivariance of deep learning models with the goal of better understanding the ways that they actually capture these concepts on a formal level. We introduce a family of invariance and equivariance metrics that allow us to quantify these properties in a way that disentangles them from other metrics such as loss or accuracy. We use our metrics to better understand the two most popular methods used to build invariance into networks, data augmentation and equivariant layers. We draw a range of conclusions about invariance and equivariance in deep learning models, ranging from whether initializing a model with pretrained weights has an effect on a trained model's invariance, to the extent to which invariance learned via training can generalize to out-of-distribution data.

In this paper we present a novel approach for identifying fraudulent papers from their titles and abstracts. The premise of the approach is that there are holes in the presentation of the approach and findings of fraudulent research papers. As an abstract is intended to highlight key features of the approach as well as important conclusions the authors seek to determine if the assumed existence of holes can be identified from analysis of abstracts alone. The data set considered is derived from papers sharing a single author with labels determined based on a formal linguistic analysis of the complete documents. To detect these logical and literary holes we utilize techniques from topological data analysis which summarizes data based on the presence of multi-dimensional, topological holes. We find that, in fact, topological features derived through a combination of techniques in natural language processing and time-series analysis allow for superior detection of the fraudulent papers than the natural language processing tools alone. Thus we conclude that the connections and holes present in the abstracts of research cons contributes to an ability to infer the scientific validity of the corresponding work.

With the recent addition of large, curated and labeled data sets to the remote sensing discipline, deep learning models have largely surpassed the performance of classical techniques. These deep models, typically Convolutional Neural Networks, are invariant to translation through the use of successive convolution layers which are themselves equivariant to translation. Further, the combination of multiple convolution and pooling layers means that in practice, the model is also approximately invariant to translation. However, until recently these models could only approach rotational invariance through data augmentation. Here we propose using a new model formulation which achieves rotational equaivariance without data augmentation for overhead imagery classification. We utilize the popular xView data set to compare the rotational equivariance formalization against a regular CNN and CNN with rotational data augmentation for the task of image classification.

We have previously shown that augmenting orthogonal matching pursuit (OMP) with an additional step in the identification stage of each pursuit iteration yields improved k-sparse reconstruction and denoising performance relative to baseline OMP. At each iteration a "path," or geodesic, is generated between the two dictionary atoms that are most correlated with the residual and from this path a new atom that has a greater correlation to the residual than either of the two bracketing atoms is selected. Here, we provide new computational results illustrating improvements in sparse coding and denoising on canonical datasets using both learned and structured dictionaries. Two methods of constructing a path are investigated for each dictionary type: the Euclidean geodesic formed by a linear combination of the two atoms and the 2-Wasserstein geodesic corresponding to the optimal transport map between the atoms. We prove here the existence of a higher-correlation atom in the Euclidean case under assumptions on the two bracketing atoms and introduce algorithmic modifications to improve the likelihood that the bracketing atoms meet those conditions. Although we demonstrate our augmentation on OMP alone, in general it may be applied to any reconstruction algorithm that relies on the selection and sorting of high-similarity atoms during an analysis or identification phase.

We introduce the concept of dynamically growing a neural network during training. In particular, an untrainable deep network starts as a trainable shallow network and newly added layers are slowly, organically added during training, thereby increasing the network's depth. This is accomplished by a new layer, which we call DropIn. The DropIn layer starts by passing the output from a previous layer (effectively skipping over the newly added layers), then increasingly including units from the new layers for both feedforward and backpropagation. We show that deep networks, which are untrainable with conventional methods, will converge with DropIn layers interspersed in the architecture. In addition, we demonstrate that DropIn provides regularization during training in an analogous way as dropout. Experiments are described with the MNIST dataset and various expanded LeNet architectures, CIFAR-10 dataset with its architecture expanded from 3 to 11 layers, and on the ImageNet dataset with the AlexNet architecture expanded to 13 layers and the VGG 16-layer architecture.

As new sensing modalities emerge and the presence of multiple sensors per platform becomes widespread, it is vital to develop new algorithms and techniques which can fuse this data. Previous attempts to deal with the problem of heterogeneous data integration for the applications in data classification were either highly data dependent or relied on simply fusing classifier outputs. In this paper we examine several related approaches: graph fusion, operator fusion, and feature space fusion. They are all associated with graph diffusion processes generated by appropriately designed operators. Our results do not make any assumptions about the data and can be easily extended to new additional modalities.