MARK 5016 Update – April Tags and Improved Models

It’s that time of the year again. What time, you would ask?

New MARK Update time!

Our R&D team is on track with adding new functions and capabilities to to Make A Robot Kit. Current update, 5016 brings a plethora of new features and bug fixes, including, but not limited to:

  • April Tags detection blocks
  • Display picture block
  • Increased distance of detection for Traffic Signs detection model and increased accuracy for Common objects recognition model
  • Adding Area parameter for all detection blocks

Let’s unpack these. We have discussed April Tags in our blog before, essentially they are big QR codes, that are easily detectable from large distance.

https://april.eecs.umich.edu/software/apriltag

MARK had the capability to detect April Tags before, but it was only limited to advanced users, capable of writing code in Micropython. Now we have added two new blocks to Codecraft

Try April Tags with this new example lesson, that can be opened directly in Codecraft!

As per user feedback we have added display picture block, that you can use to display images from SD card or FLASH memory.

The images must be adjusted to MARK screen resolution(240×320) before placing them to SD card or FLASH memory. Now you can make your MARK wish you or your relatives Happy Thanksgiving!

Another two improvements also come from user feedback. One of our early alpha-testers, Philip Mallon have wrote an extensive review, where he mentioned it would be great if detection distance for traffic signs is improved – this way an actual self-driving car demo can be made, where MARK would spot the traffic sign from a distance and react accordingly. Our Computer Vision engineers spent a lot of time tweaking the neural network model and here is the result, BEFORE and AFTER:

The new model can detect much smaller traffic signs – or the original bigger flashcards, but from larger distance! This by itself, wouldn’t be very useful, since you only want the car to react to traffic sign, when it gets close enough to it. This is why we added new parameter, called Area, to all detection blocks(not only traffic signs, but also numbers, circles, faces and others, including your custom models).

Area for rectangular objects is calculated as width of bounding box multiplied to height of the bounding box, in pixels. Area for circle is calculated as radius squared multiplied by Pi. You can use Area for object of known size to infer the information about distance. You can try Area parameter with this new example lesson.

Additionally, there are following minor changes:

  • in case of multiple detections model inference should return the result with biggest area of ​​detection
  • minor interfaces changes in Codecraft(knapsack->backpack, smartphone->smartphone, etc)
  • bug fix for temp&humidity sensor outputting string if timeout
  • fix for servo not accepting 180 degrees angle

If you still haven’t got MARK, you can now purchase it on Amazon or contact our sales team at contact@chaihuo.org for bulk orders. Stay tuned for more exciting updates next month!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s