Computer Vision Capabilities to Support Army Initiatives RFI.docx

SPECIAL NOTICE: Computer Vision Capabilities to Support Army Initiatives.

1. INTRODUCTION

REQUEST FOR INFORMATION (RFI). The purpose of this posting is to notify industry of a Request for Information (RFI). Project Manager Intelligence & Analytics (PM IS&A) within Program Executive Office Intelligence, Electronic Warfare and Sensors (PEO IEW&S) is issuing this RFI as a means of conducting market research to help scope a future contract activity, in partnership with Chief Digital and AI Office (CDAO), to train and deliver models to Army programs.

Project Linchpin enables the Army to deliver trusted solutions for AI-enabled systems through a secure Artificial Intelligence/Machine Learning Operations (AI/MLOps) environment and services while maximizing the Department of Defense (DoD) and Intelligence Community (IC) investments. Project Linchpin creates a collaborative yet competitive environment of industry and government partners to establish the AI/MLOps ecosystem and creates new opportunities for inclusion into the ecosystem through rapid agile acquisition/contracting approaches.

2. DISCLAIMER

THIS IS NOT A NOTICE OF RFP ISSUANCE. Responses to this request are voluntary and will not affect any contractor's ability to submit a proposal if, or when, an RFP is released for a future effort. The requested information is for planning purposes only and does NOT constitute a commitment, implied or otherwise, that a procurement action will be issued. It is issued to invite industry to share their thoughts with the Government on training AI Computer Vision models for Army Programs.

No entitlement to payment by the Government of direct or indirect costs or charges will arise as a result of the submission of information in response to this request.

Information received will not be considered confidential/proprietary unless marked accordingly. Company proprietary data will be reviewed by Government employees and authorized representatives only. The Government shall not be liable for or suffer any consequential damages for any improperly identified proprietary information. No award will be made as a result of this request. This request is for informational purposes only to promote exchanges of information with industry in accordance with FAR part 15.201. The Government will not pay for any information and/or materials received in response to this request for information and is in no way obligated by the information received.

3. RESPONSES

Responses related to this Government issued RFI are due no later than close of business 1 December 2023

Responses and any potential questions shall be sent via email address to: usarmy.apg.peo-iews.mbx.linchpin@army.mil; Subject Line: “Project Linchpin RFI Computer Vision Capabilities to Support Army Initiatives RFI #4”.

Responses to this RFI should be limited to 7 pages or less.

Please include the following information in all responses.

· Company Size (Small/Large)

· Defense Contractor (Traditional/ Non – Traditional)

4. BACKGROUND

The Army’s pivot to Large-Scale Combat Operations (LSCO) has underscored a critical need for improved capability to identify, monitor, and strike targets from farther distances with enhanced precision and with sharply reduced sensor-to-shooter timelines. Deep sensing is a crucial enabler for this capability. The Army also requires enhanced Line-of-Sight sensor to shooter capabilities. To address these capability needs, Project Linchpin is requesting information on object detection and computer vision capabilities to support Army initiatives.

Specifically, Project Linchpin is interested in computer vision capabilities to detect and classify objects using the following controlled unclassified information (CUI) labeled datasets:

· Horizontal Full Motion Video (FMV) Electro Optical (EO) and Infrared (IR) originating from sensors mounted on ground combat vehicles e.g. 2nd / 3rd Gen Forward Looking Infrared (FLIR).

· Overhead EO, IR and Synthetic Aperture Radar (SAR) imagery originating from overhead satellites e.g. BlackSky, WorldView.

In the context of this RFI, it is assumed that the Government will offer the following support:

· Representative sensor data along with labeled sensor data for training and testing purposes. The government will perform 3rd party validation using a sequestered data set.

· A secure hosting environment for model training that utilizes hybrid computing resources.

· Minimum set of AI/ML frameworks and libraries with the option to onboard vendor best of breed products, solutions, service to the maximum extent possible.

Note: The scope is not solely limited to supervised learning models. A variety of techniques (e.g. supervised, semi-supervised, unsupervised learning) to meet customer's AI/ML needs are encouraged and should be considered.

5. QUESTIONS TO INDUSTRY

Using the two data modalities mentioned above (Horizontal FMV / Overhead Imagery) and the assumptions, Project Linchpin is interested in industry feedback on the following questions:

1. Model training approach

a. What is your company’s approach to training and fine-tuning models? Please include any desired solutions, products, services, and timeline to produce an initial model for operational feedback.

b. Describe any base models you plan to start with and the original training set for the models.

c. What is the minimum and recommended computing resources required to train a model using an enterprise approach; and to run model inference in a tactical environment? Make reasonable assumptions about the model parameters and model architecture and provide those assumptions with your response.

2. Team Structure

a. What skillsets and team structure are required to deliver trained models? Please include quantity / roles and responsibilities for the team and overhead.

3. Cost, Contracting, and Intellectual Property (IP)

a. What is the annual cost and business model to deliver trained models?

b. What is your IP strategy, understanding that models will be continuously trained on Government data? Please include details on what the government owns vs. what industry owns.

c. How should the Government contract for model training capabilities and services?