Develop innovative edge AI applications using models trained on Edge Impulse or open-source licensed models. Bring your ideas to life by deploying them on edge AI hardware such as Arduino, ESP32, Rubik Pi, Raspberry Pi, or Android devices, and more—showcasing creativity and problem-solving in action.

Submission Guidelines

In this track, your focus must be on the application, leveraging one or several Edge Impulse models.

• Projects must integrate at least one Edge Impulse model into a working application deployed to edge hardware (e.g. Raspberry Pi, STM32, ESP32, Arduino, Android app, and more)
• Model usage:
  • Participants can train a model using Edge Impulse or use existing models.
  • In case of using existing models trained outside of Edge Impulse, the participants must use the Bring Your Own Model (BYOM) feature to integrate them into an Edge Impulse project.
  • In any case, the datasets and models used must be open-source with permissive licenses, including commercial use.
  • In case you are using public datasets add the proper attribution and citation in your documentation.
• Application innovation:
  • The focus should be on the innovation and practicality of the edge AI application developed using Edge Impulse Studio.
  • The application should solve a real-world problem or demonstrate a compelling proof-of-concept.
  • Feel free to suggest creative solutions (e.g. low-power optimizations, sensor fusion, LLM and real-time inference, and others).

View mandatory submission guidelines

Harness the power of Edge Impulse Studio to design and optimize machine learning models for audio, vision, and sensor-based data—with a strong focus on innovation, accuracy, and performance. While open-source datasets can be used as a starting point, participants are encouraged to collect or generate their own data to tailor solutions for real-world use cases.

The ultimate goal: build models that not only perform well but also run efficiently on Edge AI devices.

Submission guidelines

In this track, your focus must be on the dataset and model accuracy.

• Projects must include at least one trained ML model built using Edge Impulse Studio
• Dataset quality:
  • Participants must use high-quality datasets for training their models
  • Datasets should be well-documented and relevant to the use case
  • The dataset must be open-source with permissive licenses, including commercial use
  • Participants are encouraged to collect their own datasets
• Use case:
  • Models should be designed for a specific use case, demonstrating either generalization or a scoped application
  • Models can target classification, anomaly detection, object detection, regression, etc. using data from audio, vision or sensors in Edge Impulse Studio platform
• Iterations and research:
  • Participants should document their iterative process, including experiments, optimizations, and research conducted to improve model performance
  • Participants should demonstrate how the model was validated
  • A clear explanation of the model pipeline, dataset handling, and rationale behind model choices should be included in the submission

View mandatory submission guidelines