Beyond data capture lies true understanding. Building on our foundation of precise data acquisition, DJI Enterprise is now empowering drones to analyze and interpret the world they see. From the intelligent Matrice 4 Series to the Manifold 3 onboard computing power box, we are building AI operational platforms that bridge the gap between perception and decisive action, expanding the frontiers of productivity.
The DJI Enterprise Drone Onboard AI Challenge invites you to co-create the future of aerial intelligence. Evaluated by esteemed experts and supported with prizes and official benefits, this challenge is designed to move the strongest solutions from code to real-world operations. Together, we will redefine the era of AI-enabled flight.
March 2026
Submissions Open
Submit by May 10, 2026
May 2026
Qualified Solutions Announced
Public Testing & Voting Begins
June 2026
Public Testing & Voting Closes
June 2026
Winners Announced
Open worldwide to enterprises, research institutions, universities, developer teams, and individuals with relevant technical capabilities.
* Employees of DJI and its affiliates other than authorized dealers are not eligible to participate. DJI reserves the right of final interpretation regarding entry eligibility.
Participating solutions must be developed based on the DJI Matrice 4 Series, Matrice 4D Series, Dock 3 (including the Matrice 4D Series), Matrice 400, or Manifold 3 platforms, and must feature an onboard AI algorithm that can be deployed on the aforementioned drones or Manifold 3 platform.
The Challenge brings together 10 distinguished experts from the AI and drone industries.
* Judges are listed in no particular order.
Zhang Chun
Director, Institute of Integrated Circuit Design and Electronic Design Automation, School of Integrated Circuits, Tsinghua University
Xiong Shengwu
Director, Hubei Innovation Academy for Diverse Computing Applications
Xia Guisong
Executive Dean, School of Artificial Intelligence, Wuhan University; Deputy Director, National Engineering Research Center for Multimedia Software
Yang Ou
Dean, School of Artificial Intelligence, Shenzhen Polytechnic University
Yi Shuhua
Professor (Cuiying Scholar) and Ph.D. Supervisor, Lanzhou University
Liu Gao
Director, Technology Supervision Department, Electric Power Research Institute of Guangdong Power Grid Company
Guo Xunzhuo
Deputy Director, Navigation Management Division, Nanjing Maritime Safety Administration
Li Renzhi
Assistant Research Fellow, National Institute of Natural Hazards, Ministry of Emergency Management
Huang Ming
Professor-Level Senior Engineer, Liaocheng Natural Resources and Planning Bureau
Zhou Yuan
Director, Ecosystem Big Data R&D Center, Northeast Forestry University

Best Onboard AI Model Award
5 Winners
Prize Package
Osmo Pocket Series Combo
Manifold 3 Combo
Trophy

Industry Application Excellence Award
10 Winners
Prize Package
Manifold 3 Combo
Trophy
Global Exposure
Solutions will be promoted on DJI Enterprise's official global channels, reaching industry clients worldwide.
Catalog Inclusion
Solutions will be featured in the official Onboard AI Solutions Catalog, showcased to DJI's entire customer base.
Partner Fast-Track
Teams can gain fast-track access to the DJI Enterprise Ecosystem partner audit, with opportunities for priority access to benefits such as new product beta testing and direct technical support.

Finalist Award
Multiple Winners
Prize
Award Certificate
DJI opens onboard computing capabilities to developers. With only a few steps for model training and conversion, your algorithms can run natively on the drone, enabling intelligent detection tailored to real-world operational scenarios.
AI algorithms deployed on Matrice 4T have supported rhino protection efforts in the Sabi Sand Nature Reserve. Combining onboard thermal imaging and long-range zoom with AI-based detection, the system helps the local Reaction Team respond more effectively to poaching threats.
Matrice 400, paired with Manifold 3, enhances inspection efficiency. Developers can deploy AI detection algorithms on Manifold 3, enabling the system to accurately detect floating debris, aquatic vegetation overgrowth, and silt accumulation in real time, ultimately supporting ecological assessment workflows.
Manifold 3 packs an impressive 100 TOPS of computing power into a compact unit, weighing about 120 g. When mounted onto the Matrice 400 and Matrice 4 Series drone platforms, it provides enhanced computing power for flight, gimbal, and camera control, optimizing efficiency in various operations. With a user-friendly development environment, Manifold 3 empowers developers to build high-performance applications across diverse scenarios, from energy inspections to public safety and beyond.

With DJI FlightHub 2's high-definition, low-latency video streaming, developers can seamlessly connect live drone feeds to their own cloud-based AI inference services. Detection results can be viewed remotely in real time, while automatic alerts support faster decision-making, making drone operations smarter and more efficient.

Yes. While early pre-registration was opened through certain channels in February, you can still submit your solution directly on the official Challenge website even if you did not participate in that phase.
The deadline for all submissions is May 10, 2026.
Yes, you can. If you submit multiple versions of the same solution, we will only evaluate the most recently submitted version as your final entry.
Yes. Any solution involving any of the following will be disqualified from the Challenge:
1. Violating any applicable local or national laws and regulations.
2. Infringing upon the lawful rights and interests of others, including but not limited to privacy rights and copyrights.
3. Containing content that is discriminatory, offensive, or contrary to public order and moral standards.
4. Involving any other severe misconduct or inappropriate circumstances, as determined by the Organizing Committee.
Yes. You may submit multiple solutions, but a team is limited to a maximum of 3 shortlisted solutions in total, and only 1 per category (Refer to the solution submission form on the official website for detailed category definitions). If you submit more than 3 qualifying submissions, the Organizing Committee will select the top 3 based on evaluation criteria.
Your submission must include the following:
a. Solution Proposal Document: A complete solution overview in PDF or PPT format.
b. Onboard Model File: Must be configured for the Designated Test Device SN (authorization valid through July 30, 2026).
c. Validation Dataset: Minimum of 20 test sample images.
d. Full Operation Screen Recording: Must capture the complete workflow from model import to report generation.
e. IP Declaration.
Due to flight safety considerations, solutions submitted under the Manifold 3 AI Algorithms category are excluded from public trials and authorized device count statistics. Consequently, for the Authorized Device Count dimension (which has a maximum of 25 points):
a. Manifold 3 solutions are awarded a fixed score of 15 points, which serves as the maximum achievable score for these solutions in this dimension.
b. They are excluded from ranking-based bonus points.
As a guideline, no more than five Manifold 3 solutions will advance to the Finals.
No, actual commercial results are not required. Your model qualifies as long as it can effectively detect the target object.
You (the developer) retain full intellectual property rights to your model. Additionally:
For models and algorithms, by submitting a solution, you authorize DJI and its affiliates to use them for Challenge evaluation and testing. This testing authorization is valid until July 30, 2026.
For solution overviews, demonstration videos, and other related materials, DJI and its affiliates will publish these on DJI's official website, social media, or other promotional channels after obtaining your authorization.