Exploring Smart Eyes on MCU -Join Seeed Vision Challenge to win BIG!

Background of this Challenge

The AI Vision sensor is a sensor that we believe has great potential ability to help you see and analyze scenes that were previously unseen. Here’s why:

  1. It is a viable replacement for many traditional sensors: For instance, PIR Sensors for motion detection, Ultrasonic Sensors for measuring distances, Thermopiles and IR Sensors for temperature readings, and Contact Sensors for detecting touch or proximity—all have their unique features. However, vision sensors excel in precision detection. While traditional sensors might guess if someone has fallen, a vision sensor can confirm the event accurately.
  2. Unparalleled flexibility: Equipped with the right AI model, a vision sensor can act as a termite detector, a snake sensor, or even spot blockages at fire exits—tasks that previously required complex IP cameras or regular manual checks. AI vision sensors bring a level of adaptability absent in their traditional counterparts.

However, earlier MCU-based vision sensors were limited by computing power, often only capable of simple applications like face detection. We think the Arm Cortex M55 & Ethos U55 represents a game-changer here, as it improves machine learning performance by 480 times compared to existing systems based on Cortex-M architecture.

Seeed recently developed a vision sensor module, the Grove Vision AI Module v2, which is based on the Himax WiseEye2 HX6538 that utilizes this ARM architecture.

We have conducted some tests, and the process and results can be found here. The results are quite surprising—it can even be used for pose detection, which is a truly impressive feat for MCU performance. We believe it opens up functionalities previously inaccessible to traditional vision sensors.

Compared to previous vision sensors, it has three significant advantages:

🌟Significant uplift in ML capabilities

🌟Award-winning low power consumption, making it suitable for battery-powered applications.

🌟No-code model deployment with Seeed Studio SenseCraft AI, which has an MCU-optimized model library for peak performance that is easily accessible.

With such an exciting product, we invite you to join in the fun. Seeed, in collaboration with Arm and Himax, is launching a Vision Challenge, and we’re eager to witness your innovative use of this product.

Steps to participate in this Vision Challenge

Step 1:
Submit your ideas

Tell us what you’d like to create or the real-world problems you aim to tackle using the Grove Vision AI V2 Sensor. We’ll review all submissions and select the 250 most innovative ideas. Chosen participants will receive a complimentary Grove Vision AI Module v2 and a compatible OV5647 camera module to bring their concepts to life.

Seize this unique opportunity by submitting your ideas now—availability is on a first-come, first-serve basis.

After you submit your idea, we will review it, and if your idea is selected, we will contact you via the email address you provided and ship you the products. Should you have any questions, please feel free to reach out to iot[at]seeed.cc, and we will be there to support you. Please note that we’re not accepting project ideas any more, please join the challenge in step 2 by submitting your projects by June 30, 2024.

Step 2: Publish your content and win

Showcase your work by June 30, 2024 on platforms like Hackster, Instructables, or YouTube or any others, and then share the project link with us by clicking the button below.

We’re open to all content but consider these directions:

  • Conduct diverse evaluations to test its capabilities. Try running models like Mobilenet V1, V2, Efficientnet-lite, Yolo v5 & v8 in TensorFlow and PyTorch to test its performance
  • Apply it to your own applications and solve your challenges by training your own models or using our pre-trained model library.
  • Share your user experience by integrating it with popular development boards like Raspberry Pi or ESP32. Let us know how it performs as a subprocessor for image processing, and whether or not it meets your expectations
  • Impress us with your creativity.

Here are some tutorials from which you may get inspiration:

1. Some Demo Showcase

3 cool demos that our team created: automatic lock, gesture-controlled mouse, face-tracking fan.

2. Train Your Models

Apart from the available models on our SenseCraft AI Platform for no-code deployment, you can also train your custom models. Refer to this wiki.

3. Home Assistant

Want to build a smart home application? You can push all the data/results of the Grove Vision AI Module to Home Assistant.

4. Telegram Usage

You can also Deploy Grove Vision AI V2 To Notify You Via Telegram.

5. Intelligent IP Camera With AI Function

As well as turn the Grove Vision AI into a smart IP camera via a HTTP server.

6. Bird Feeder Defender

The surveillance system made by skruglewicz was designed to prevent squirrels from invading my bird feeders. More about this project.

7. Drunks Detection

Drunks are detected with AI, reported with BLE, and dispersed with dog barks. Learn more about the project made by Roni Bandini.

8. More Community Projects

If you want to explore more projects from the community, here is a collection of the projects.

✨Before the end of June, we will select outstanding content from all submissions, and the exceptional creators will receive the following benefits:✨

1)3 Top Picks – Our favorite 3 publications will each receive a $300 USD coupon to purchase any Seeed products.
2) Alpha test our to-be-released SenseCAP Watcher (picture below)- For those whose work involves Computer Vision, TinyML, or LLM, we’ll select participants to join our alpha test of this new product—with no limit on the number of selections, just great work.
3) Official Promotions – Exceptional content may be featured through official Seeed, Himax, and ARM channels, like social media or trade shows.
4) Co-Creation Opportunity – Your project could be selected of our Seeed-Co-Create program, potentially turning your work into a product with a significant royalty for every unit sold.


📅Submit your content before 30 June, 2024

We can’t wait to see your creativity, join our Vision Challenge NOW

Should you have any questions, please feel free to reach out to iot[at]seeed.cc, and we will be there to support you.

Subscribe to our Newsletter for Exclusive Updates!

About Author

1 thought on “Exploring Smart Eyes on MCU -Join Seeed Vision Challenge to win BIG!

  1. The ability to detect obstructions in fire exits on each floor without the need for complicated IP cameras or regular patrols is a significant advancement in terms of safety and efficiency.

Comments are closed.


March 2024