Award Data

In this proposed Phase II effort we will implement a software framework that will reduce the time required to annotate large image/video datasets by a factor of 100x while also reducing the data necessary to train state of the art computer vision models by up to 80%. Our strategy combines several elements and ideas. First, we have developed a sub-tile based a priori theory of how and why CNNs can ...

In Phase I we developed and demonstrated a bi-directional, monolithic, High-Speed, Electro-optic shutter for Range Gated Imaging. While meeting the majority of the key performance parameters for Range-Gated Imaging at high Pulse Repetition Frequency, improvements can be achieved in transmission, contrast and reliability. The Phase Ii activity begins with a upgraded Electro-optic Shutter incorpora ...

Impact Observatory recently demonstrated production of the world’s first fully automated time series of annual global land use land cover (LULC) maps at 10m resolution using deep learning algorithms applied to Sentinel-2 satellite imagery. We propose to develop an AI-powered prototype system for automated mapping and change monitoring of the world that can be applied continuously, in near real-t ...

TP Engineering personnel have extensive experience with electro-optic systems and high Pulse Repetition Frequency (PRF) Laser systems. We have detailed knowledge of Pockels cell systems enabling active gated imaging through foliage at PRF 100 kHz PRF. Such systems can dramatically improve and protect Geiger-mode LIDAR by both controlling the transmitter output and gating out unwanted return lig ...

Deep Convolutional Neural Networks (DCNNs) have become ubiquitous in the analysis of large datasets with geometric symmetries. These datasets are common in medicine, science, intelligence, autonomous driving and industry. While analysis based on DCNNs have proven powerful, uncertainty estimation for such analyses has required sophisticated empirical studies. This has negatively impacted the effect ...

In the pattern of life problem space, data is often represented via mathematical graphs, in which a variety of algorithms may be employed to conduct semi-autonomous analysis. While successful empirical application of graph-domain algorithms on ABI problems has been achieved, most of these algorithms require a tuning parameter, which is often set heuristically in real-world scenarios. Arete has dev ...

In Phase II we will explore further what processing of synthetic image generation produces training data that optimizes the model’s performance on real world images. While existing work on CycleGAN optimizes for human visual perception of domain adaptation it may not necessarily be optimal for training a network to generalize to real scenarios. We will look at modifications of the loss functions ...

In its effort to provide necessary intelligence and analysis, the National Geospatial-Intelligence Agency (NGA) utilizes extensive traffic camera systems. However, the large amount of data overwhelms both analysts and existing processing methods. In order to provide a better understanding and reduce the search space for common problems such as target tracking, it is necessary to extract the camera ...

Arete will integrate the Mover Intelligence Extraction Engine with the Social Structured Framework (SSF) to produce a capability to generate narratives of MOVINT track data with context derived from associated Geographic Information Systems (GIS) and other foundation data. The Extraction Engine will be derived from the Arete Cognitive Behavior Classifier (CBC) to characterize MOVINT tracks into co ...

The team of Signature Research, Inc. and Michigan Technological University will develop and demonstrate methods and metrics to evaluate the performance of machine learning-based computer vision algorithms with low numbers of samples of labeled EO imagery. We will use the existing xView panchromatic dataset to demonstrate a proof-of-concept set of tools. If successful, in Phase II, we will extend t ...