(Photo Credit: U.S. Army) VIEW ORIGINAL

The Office of Business Transformation continues to work alongside the U.S. Army Combat Capabilities Development Command Analysis Center Army Research Laboratory (ARL) to facilitate the second phase of the Deep Green Data Science and Artificial Intelligence competition, which will start on January 9, 2023.

Phase II of the Autonomous All-Terrain Vehicle Semantic Image Segmentation Competition looks to further develop a Deep Learning approach for non-urban autonomous environment understanding.

Through the application of computer vision and LiDAR systems, the Deep Green competition will be focused on developing the initial launching point for further state-of-the-art Deep Learning research. The goal is to develop a model that will allow an autonomous all-terrain vehicle to navigate through rugged all-terrain environments in support of the Warfighter.

According to Mr. Bakari Dale, the Army’s Senior Advisor for Enterprise Data Science & Artificial Intelligence, "the Army is also looking to recruit new talent into its digital workforce, and has opened up Phase II of the competition to the Service Academies, Senior Military Colleges, and prestigious academic institutions." Mr. Dale is excited about the innovative AI tools and data science up-skilling that the participants will have the opportunity to experience.

Pre-Registration for Phase II is now open. If interested, please contact Dr. Matthew Millar (see below) for more information and registering for the competition. Phase II takes place from January 9 through July 28, 2023, and registration will remain open until March 1st.

Phase II of the competition will consist of two tracks for grading: the Computer Vision track, and the LiDAR sensor track. A competitor may choose one or both datasets to build out their solution, and may also enter one or both tracks of the competition.

The Computer Vision track allows for both Python-PyTorch/Tensorflow or H2O.ai Hydrogen Torch models to be submitted. The LiDAR track will only support Python-PyTorch/Tensorflow models. Training resources and instruction in Computer Vision, LiDAR, Python, Pytorch/Tensorflow, and H2O.ai Hydrogen Torch tools will be provided to competitors, based on what track(s) the participate in.

The working environment will be fully remote, allowing for fully distributed teams to work together.

To enable even greater collaboration across multiple individuals and smaller teams, a roster with group numbers will allow for members to register to join a group. Individuals may also work by themselves. A signup sheet for groups will be posted in the Deep Green Phase II Teams' channel (see link below) on 9 January, which will allow for registered competitors, without a team, to sign up to and be part of a distributed team/group.

For Phase II of the competition, the model with the highest evaluation metric score will be used as a baseline performance prototype for further development on Army autonomous vehicle systems. The computer vision and LiDAR tracks will have separate ranking leaderboards, with the overall first place model (highest score) used in the development of advanced autonomous robotics systems supporting warfighter security and platoon operations.

We look forward to the innovative approaches and developments that will come from Phase II of the 2022-2023 Deep Green Data Science and Artificial Intelligence Competition.

Deep Green Autonomous All-Terrain Vehicle Phase II Teams Channel

POC: Dr Matthew Millar: matthew.c.millar3.civ@army.mil