Products
ACES
CoVar are the world’s experts in building state-of-the-art AI/ML pipelines for automatic target recognition (ATR). Our ACES software is a multi-tiered ATR solution, with custom, patented algorithms capable of detecting tiny unresolved distant targets, mid-range objects, and counting wheels, identifying weak spots, and providing targeting information for closer targets. ACES works on IR (LWIR, MWIR, SWIR, NIR) and RGB images, runs on low-power FPGA’s, Qualcomm Devices, NVIDIA edge devices, or in the cloud. ACES can output visual cues, CoT, or TAK messages. ACES has been tested in ground-to-ground, air-to-ground, surface-to-surface, and surface-to-air applications. ACES is highly configurable to your ATR needs; we have specialized implementations targeting varying levels of video compression, bandwidth considerations and latency/throughput/power considerations.
ACES Demonstration videos are available to government customers on https://www.tradewindai.com/
Example ACES capabilities showing part-based target decomposition and classification.
Contact us about pricing and detailed capabilities.
Multiple ACES Publications are available on DTIC – e.g., see:
Crosskey, M., Mandal, S., Milner, M., Talbott, M., Torrione, P., Confidence Jitter in Convolutional Neural Networks for AiTR: Impacts and Mitigations. Military Sensing Symposium, 2022.
Crosskey, M., Mandal, S., Reynolds, J., Talbott, M., Torrione, P., Part-Based Identification of Vehicles in IR Imagery. Military Sensing Symposium, 2022.
Brandon, N., Talbott, M., Torrione, P., Detecting Small Moving Targets via DeepMTI (Deep Moving Target Indicators). Military Sensing Symposium, 2023.
Tran, M., Viola, A., Talbott, M., Torrione, P., Morton, K., LW/MW Sensor Fusion for Improved AiTR. Military Sensing Symposium, 2023.
Camilo, J., Talbott, M., Torrione, P., Recent Developments in Deploying Convolutional Neural Networks on Low Power FPGA Hardware. Military Sensing Symposium, 2023.
Viola, A., Talbott, M., Torrione, P., 3-D World Space Tracking by Target Segmentation. Military Sensing Symposium, 2023.
Tran, M., Mandal, S., Morton, K., Torrione, P., Viola, A., Talbott, M., LW/MW Sensor Fusion for Improved AiTR, NATO MSS 2023.
Tran, M., Torrione, P., Hull, P., Camilo, J, Sun, Yi., Hutchison, J., Talbott, M., AITR Power, Frame Rate, and Network Complexity on Relevant FPGAs, Military Sensing Symposium, 2024.
Spell, G., Viola, A., Torrione, P., Talbott, M., Improving AiTR Algorithms with Infrared and RGB Fusion, Military Sensing Symposium, 2024.
S. Mandal, A. Kompella, A. Viola, P. Torrione, J. Miller, M. Talbott, On Cross-Camera AiTDR Generalizability, Military Sensing Symposium, 2025.
S. Gupte, A. Viola, S. Mandal, J. Jimenez, M. Talbott, Rapid Rendering and Integration of Digital Elevation Information in AiTDR Pipelines, Military Sensing Symposium, 2025.
N. Brandon, P. Torrione, M. Talbott, Limits of Detectability for Low Contrast Moving Targets, Military Sensing Symposium, 2025.
CoVar are the world’s experts in building state-of-the-art AI/ML pipelines for automatic target recognition (ATR). Our ACES software is a multi-tiered ATR solution, with custom, patented algorithms capable of detecting tiny unresolved distant targets, mid-range objects, and counting wheels, identifying weak spots, and providing targeting information for closer targets. ACES works on IR (LWIR, MWIR, SWIR, NIR) and RGB images, runs on low-power FPGA’s, Qualcomm Devices, NVIDIA edge devices, or in the cloud. ACES can output visual cues, CoT, or TAK messages. ACES has been tested in ground-to-ground, air-to-ground, surface-to-surface, and surface-to-air applications. ACES is highly configurable to your ATR needs; we have specialized implementations targeting varying levels of video compression, bandwidth considerations and latency/throughput/power considerations.
ACES Demonstration videos are available to government customers on https://www.tradewindai.com/
Example ACES capabilities showing part-based target decomposition and classification.
Contact us about pricing and detailed capabilities.
Multiple ACES Publications are available on DTIC – e.g., see:
Crosskey, M., Mandal, S., Milner, M., Talbott, M., Torrione, P., Confidence Jitter in Convolutional Neural Networks for AiTR: Impacts and Mitigations. Military Sensing Symposium, 2022.
Crosskey, M., Mandal, S., Reynolds, J., Talbott, M., Torrione, P., Part-Based Identification of Vehicles in IR Imagery. Military Sensing Symposium, 2022.
Brandon, N., Talbott, M., Torrione, P., Detecting Small Moving Targets via DeepMTI (Deep Moving Target Indicators). Military Sensing Symposium, 2023.
Tran, M., Viola, A., Talbott, M., Torrione, P., Morton, K., LW/MW Sensor Fusion for Improved AiTR. Military Sensing Symposium, 2023.
Camilo, J., Talbott, M., Torrione, P., Recent Developments in Deploying Convolutional Neural Networks on Low Power FPGA Hardware. Military Sensing Symposium, 2023.
Viola, A., Talbott, M., Torrione, P., 3-D World Space Tracking by Target Segmentation. Military Sensing Symposium, 2023.
Tran, M., Mandal, S., Morton, K., Torrione, P., Viola, A., Talbott, M., LW/MW Sensor Fusion for Improved AiTR, NATO MSS 2023.
Tran, M., Torrione, P., Hull, P., Camilo, J, Sun, Yi., Hutchison, J., Talbott, M., AITR Power, Frame Rate, and Network Complexity on Relevant FPGAs, Military Sensing Symposium, 2024.
Spell, G., Viola, A., Torrione, P., Talbott, M., Improving AiTR Algorithms with Infrared and RGB Fusion, Military Sensing Symposium, 2024.
S. Mandal, A. Kompella, A. Viola, P. Torrione, J. Miller, M. Talbott, On Cross-Camera AiTDR Generalizability, Military Sensing Symposium, 2025.
S. Gupte, A. Viola, S. Mandal, J. Jimenez, M. Talbott, Rapid Rendering and Integration of Digital Elevation Information in AiTDR Pipelines, Military Sensing Symposium, 2025.
N. Brandon, P. Torrione, M. Talbott, Limits of Detectability for Low Contrast Moving Targets, Military Sensing Symposium, 2025.
BULLSEYE
There is a need to provide general purpose advanced AI/ML so every soldier can use attritable unmanned aerial systems (UAS) with low-cost AI/ML devices to achieve any mission. CoVar’s BULLSEYE puts the power of AI/ML onto any drone with low-power, low-cost Android or Nett-Warrior devices that soldier’s are already carrying. CoVar’s BULLSEYE is highly configurable, and empowers users to achieve any mission on any drone. See our demonstration video here –
Left – pre-mission phone configuration. Center: Android device on drone executing mission. Right: Target reporting back to TAK/CoT.
Contact us about pricing and detailed capabilities.
BULLSEYE Publications are available on DTIC – e.g., see:
D. Kalika, F. Haltmayer, A. Viola, P. Torrione, Universal UAS Payloads: Android Devices for AI/ML Pipelines, Military Sensing Symposium, 2025.
There is a need to provide general purpose advanced AI/ML so every soldier can use attritable unmanned aerial systems (UAS) with low-cost AI/ML devices to achieve any mission. CoVar’s BULLSEYE puts the power of AI/ML onto any drone with low-power, low-cost Android or Nett-Warrior devices that soldier’s are already carrying. CoVar’s BULLSEYE is highly configurable, and empowers users to achieve any mission on any drone. See our demonstration video here –
Left – pre-mission phone configuration. Center: Android device on drone executing mission. Right: Target reporting back to TAK/CoT.
Contact us about pricing and detailed capabilities.
BULLSEYE Publications are available on DTIC – e.g., see:
D. Kalika, F. Haltmayer, A. Viola, P. Torrione, Universal UAS Payloads: Android Devices for AI/ML Pipelines, Military Sensing Symposium, 2025.
ACES HIGH & SAVANT
CoVar builds advanced generative AI tools to ingest and understand overhead satellite data. CoVar’s ACES HIGH is an extension of our ACES pipelines to processing overhead data, and can be configured and trained to ingest SAR, EO, IR images and/or overhead video.
CoVar’s SAVANT is an extension of ACES HIGH specifically designed to maintain custody on high-value targets across multiple distinct modalities (e.g., detect in EO, re-acquire in SAR).
CoVar SAVANT. Left: Pan-sharpened image data. Center: CoVar SAVANT CAD-Model Estimated from image. Right: Expected target SAR response for re-acquisition.
Contact us about pricing and detailed capabilities.
ACES HIGH Publications are available on DTIC:
Crosskey, M., Jimenez, J., Talbott, M., Torrione, P., Space and Ground Asset Fusion for Improved ATR. Military Sensing Symposium, 2023.
LeBlanc, P., Fox, C., Hull, P., Luzi, F., Camilo, J., Torrione, P., Morton, K., Jiminez, J., A Silhouette-based Approach for Classifying Objects in Overhead Imagery, Military Sensing Symposium, 2024.
P. LeBlanc, M. Crosskey, C. Fox, P. Hull, B. Purcell, J. Camilo, P. Torrione, K. Morton, J. Brosovich, J. Jimenez, Savant: One-shot CAD Model Learning in Remote Sensing with Domain Transfer, Military Sensing Symposium, 2025.
CoVar builds advanced generative AI tools to ingest and understand overhead satellite data. CoVar’s ACES HIGH is an extension of our ACES pipelines to processing overhead data, and can be configured and trained to ingest SAR, EO, IR images and/or overhead video.
CoVar’s SAVANT is an extension of ACES HIGH specifically designed to maintain custody on high-value targets across multiple distinct modalities (e.g., detect in EO, re-acquire in SAR).
CoVar SAVANT. Left: Pan-sharpened image data. Center: CoVar SAVANT CAD-Model Estimated from image. Right: Expected target SAR response for re-acquisition.
Contact us about pricing and detailed capabilities.
ACES HIGH Publications are available on DTIC:
Crosskey, M., Jimenez, J., Talbott, M., Torrione, P., Space and Ground Asset Fusion for Improved ATR. Military Sensing Symposium, 2023.
LeBlanc, P., Fox, C., Hull, P., Luzi, F., Camilo, J., Torrione, P., Morton, K., Jiminez, J., A Silhouette-based Approach for Classifying Objects in Overhead Imagery, Military Sensing Symposium, 2024.
P. LeBlanc, M. Crosskey, C. Fox, P. Hull, B. Purcell, J. Camilo, P. Torrione, K. Morton, J. Brosovich, J. Jimenez, Savant: One-shot CAD Model Learning in Remote Sensing with Domain Transfer, Military Sensing Symposium, 2025.
CONAN & ISTARI
CoVar’s CONAN is an all-in-one AI/ML pipelining tool suitable for both real-time implementation and off-line R&D. CONAN enables AI/ML system designers to connect ATR capabilities to build custom pipelines for specific missions, run these pipelines in real-time, and control the pipelines (e.g., enabling or disabling components; adjusting sensitivities) without re-starting the processes. CONAN implements an open API so that multiple vendors can collaborate without sharing proprietary information. ISTARI is an extension to CoVar CONAN that enables knowledge sharing across CONAN components via a knowledge-graph database, enabling higher-order reasoning and AI.
CoVar CONAN Pipelines. Left: CONAN Pipeline flow-diagram visualization. AI/ML components can be trivially added/removed in the pipeline, and configured in real-time to adjust parameters, settigns, etc. Right: The pipeline running in real-time detecting UXO on an airfield.
Contact us about pricing and detailed capabilities.
CONAN/ISTARI Publications are available on DTIC:
Viola A., Mandal, S., Hull, P., Jiminez, J., Istari: A Real-Time Automated Reasoning Framework, Military Sensing Symposium, 2024.
CoVar’s CONAN is an all-in-one AI/ML pipelining tool suitable for both real-time implementation and off-line R&D. CONAN enables AI/ML system designers to connect ATR capabilities to build custom pipelines for specific missions, run these pipelines in real-time, and control the pipelines (e.g., enabling or disabling components; adjusting sensitivities) without re-starting the processes. CONAN implements an open API so that multiple vendors can collaborate without sharing proprietary information. ISTARI is an extension to CoVar CONAN that enables knowledge sharing across CONAN components via a knowledge-graph database, enabling higher-order reasoning and AI.
CoVar CONAN Pipelines. Left: CONAN Pipeline flow-diagram visualization. AI/ML components can be trivially added/removed in the pipeline, and configured in real-time to adjust parameters, settigns, etc. Right: The pipeline running in real-time detecting UXO on an airfield.
Contact us about pricing and detailed capabilities.
CONAN/ISTARI Publications are available on DTIC:
Viola A., Mandal, S., Hull, P., Jiminez, J., Istari: A Real-Time Automated Reasoning Framework, Military Sensing Symposium, 2024.
HAT/InCite
CoVar InCite is a patent-pending approach to ensuring that Generative AI models provide reasonable, accurate, and trustworthy citations and references to the exact source material used to generate LLM outputs. By peering inside the LLM’s mind, InCite enables us to build systems whose responses are truly traceable, trustworthy, governable, and valid. CoVar’s Incite is built on CoVar Hallucination Awareness Tool (HAT), which extends Incite to include general purpose hallucination identification and avoidance.
CoVar interrogation of LLM internal mental states provides traceable and explicit references to source material. This enables automatic source classification and builds user trust.
Contact us about pricing and detailed capabilities.
CoVar InCite is a patent-pending approach to ensuring that Generative AI models provide reasonable, accurate, and trustworthy citations and references to the exact source material used to generate LLM outputs. By peering inside the LLM’s mind, InCite enables us to build systems whose responses are truly traceable, trustworthy, governable, and valid. CoVar’s Incite is built on CoVar Hallucination Awareness Tool (HAT), which extends Incite to include general purpose hallucination identification and avoidance.
CoVar interrogation of LLM internal mental states provides traceable and explicit references to source material. This enables automatic source classification and builds user trust.
Contact us about pricing and detailed capabilities.
Pocket ACES
There is a sign outside PEO Soldier that says “Every ounce matters, Every bullet counts”. In the battle against enemy drones, there is a need for UAS detection and early warning devices with zero additional SWAP. CoVar’s Pocket ACES puts the power of ACES into your pocket. By integrating C-UAS AI/ML pipelines into Android Devices, every phone or Nett-Warrior device can now be used as a counter-drone early warning system.
Example C-UAS drone early detection from two Android devices (left), and target triangulation (right).
Contact us about pricing and detailed capabilities.
Pocket ACES Publications are available on DTIC:
D. Kalika, F. Haltmayer, A. Viola, P. Torrione, Universal UAS Payloads: Android Devices for AI/ML Pipelines, Military Sensing Symposium, 2025.
There is a sign outside PEO Soldier that says “Every ounce matters, Every bullet counts”. In the battle against enemy drones, there is a need for UAS detection and early warning devices with zero additional SWAP. CoVar’s Pocket ACES puts the power of ACES into your pocket. By integrating C-UAS AI/ML pipelines into Android Devices, every phone or Nett-Warrior device can now be used as a counter-drone early warning system.
Example C-UAS drone early detection from two Android devices (left), and target triangulation (right).
Contact us about pricing and detailed capabilities.
Pocket ACES Publications are available on DTIC:
D. Kalika, F. Haltmayer, A. Viola, P. Torrione, Universal UAS Payloads: Android Devices for AI/ML Pipelines, Military Sensing Symposium, 2025.
C-Suite (CVRNET)
Most computer vision AI/ML code (e.g., YOLO) is generated by universities and foreign nationals, made publicly available, then downloaded and run for by US contractors for US DOD applications. There is a need for DOD generated, validated, and controlled implementations of these networks that have not been publicly released, that have been trained on different data sets, and are therefore secure and robust to adversarial attacks. CoVar’s C-SUITE provides model implementations, training scripts, augmentations, and data interfaces that have been implemented from the ground-up to avoid possible model poisoning. C-SUITE has never been released publicly and exists at the CUI level (and higher).
Contact us about pricing and detailed capabilities.
Most computer vision AI/ML code (e.g., YOLO) is generated by universities and foreign nationals, made publicly available, then downloaded and run for by US contractors for US DOD applications. There is a need for DOD generated, validated, and controlled implementations of these networks that have not been publicly released, that have been trained on different data sets, and are therefore secure and robust to adversarial attacks. CoVar’s C-SUITE provides model implementations, training scripts, augmentations, and data interfaces that have been implemented from the ground-up to avoid possible model poisoning. C-SUITE has never been released publicly and exists at the CUI level (and higher).
Contact us about pricing and detailed capabilities.
Clarity
There is a need for lightweight data analytics and understanding tools to track AI/ML and ATR system performance, identify unexpected correlations in ATR capabilities, and identify data-drift in real-time data. CoVar’s CLARITY is a universal, configurable data analytics tool that includes support for data exploration, conditional ROC-Curve / P-R Curve generation, and exploration. CoVar’s CLARITY has been deployed to the Oil & Gas Sector, and as part of US ARMY Data Collections.
CoVar CLARITY Screen Capture analyzing publicly available data. Note the interactive data selection enables deep-dives into data sub-sets. CLARITY is designed for ATR system analytics and data understanding (briefings available at the CUI level).
Contact us about pricing and detailed capabilities.
There is a need for lightweight data analytics and understanding tools to track AI/ML and ATR system performance, identify unexpected correlations in ATR capabilities, and identify data-drift in real-time data. CoVar’s CLARITY is a universal, configurable data analytics tool that includes support for data exploration, conditional ROC-Curve / P-R Curve generation, and exploration. CoVar’s CLARITY has been deployed to the Oil & Gas Sector, and as part of US ARMY Data Collections.
CoVar CLARITY Screen Capture analyzing publicly available data. Note the interactive data selection enables deep-dives into data sub-sets. CLARITY is designed for ATR system analytics and data understanding (briefings available at the CUI level).
Contact us about pricing and detailed capabilities.
Generative AI, Agentic AI, & LLMs
CoVar is building next-generation Agentic AI to solve problems that matter to the Department of Defense. Our custom approach to Agentic AI leverages best-in-class commercial grade AI, and combines them with custom API/MPC calls to interact with DoD software, existing documents, technical manuals, doctrine, training, and best practices. By way of example, our Agentic AI automatically identifies valid landing zones for a CH-47 helicopter based on technical documents and map information. Other applications include voice-command of complicated AI/ML pipelines, and automatic interpretation of commander’s intent into technical instructions for autonomous weapons systems.
Example CoVar Agentic AI combines industrial LLMs with DOD tools and applications. In this example, our AI automatically solves the users’ query for where they can safely land their Chinook.
Contact us about pricing and detailed capabilities.
Publications are available on DTIC:
A Viola, P. Torrione, J. Camilo, J. Jimenez, M. Talbott, Large Language Models for AiTDR System Control at the Edge, Military Sensing Symposium, 2025.
C. Fox, M. Wang, G. Myers, S. Mandal, K. Morton, J. Wright, Querying the Battlefield with Natural Language using an LLM-Enabled Knowledge Graph Data Interface, Military Sensing Symposium, 2025.
CoVar is building next-generation Agentic AI to solve problems that matter to the Department of Defense. Our custom approach to Agentic AI leverages best-in-class commercial grade AI, and combines them with custom API/MPC calls to interact with DoD software, existing documents, technical manuals, doctrine, training, and best practices. By way of example, our Agentic AI automatically identifies valid landing zones for a CH-47 helicopter based on technical documents and map information. Other applications include voice-command of complicated AI/ML pipelines, and automatic interpretation of commander’s intent into technical instructions for autonomous weapons systems.
Example CoVar Agentic AI combines industrial LLMs with DOD tools and applications. In this example, our AI automatically solves the users’ query for where they can safely land their Chinook.
Contact us about pricing and detailed capabilities.
Publications are available on DTIC:
A Viola, P. Torrione, J. Camilo, J. Jimenez, M. Talbott, Large Language Models for AiTDR System Control at the Edge, Military Sensing Symposium, 2025.
C. Fox, M. Wang, G. Myers, S. Mandal, K. Morton, J. Wright, Querying the Battlefield with Natural Language using an LLM-Enabled Knowledge Graph Data Interface, Military Sensing Symposium, 2025.
CoVar Data Collection & Labeling Services
CoVar has access to airfields and sensors suitable for collecting data on UAS, ground vehicles, and other objects of interest. Our data collection hardware includes IR, RGB, RF (software defined radios), acoustics, Android phone devices, and an event based imaging system. We have multiple drone pilots, and access to flight space. Data collections can be conducted for 1-5 days.
Our labeling teams are all cleared to process and handle CUI data, and our specialized network infrastructure allows us to distribute labeling jobs amongst US personnel across the continental United States. Our formal verification processes ensure that labels are accurate and precise.
CoVar data collection team at a recent data collection.
CoVar’s labelers make use of the CVAT labeling software, which CoVar deploys on a CUI-compliant network to labelers distributed across the US. Our label validation process ensures high quality labels.
Contact us about pricing and detailed capabilities.
CoVar has access to airfields and sensors suitable for collecting data on UAS, ground vehicles, and other objects of interest. Our data collection hardware includes IR, RGB, RF (software defined radios), acoustics, Android phone devices, and an event based imaging system. We have multiple drone pilots, and access to flight space. Data collections can be conducted for 1-5 days.
Our labeling teams are all cleared to process and handle CUI data, and our specialized network infrastructure allows us to distribute labeling jobs amongst US personnel across the continental United States. Our formal verification processes ensure that labels are accurate and precise.
CoVar data collection team at a recent data collection.
CoVar’s labelers make use of the CVAT labeling software, which CoVar deploys on a CUI-compliant network to labelers distributed across the US. Our label validation process ensures high quality labels.
Contact us about pricing and detailed capabilities.