Microscopes
Look through the microscope and process the images with Oracle to see if they can guess what you're looking at. If Oracle is wrong, teach them so they know more in the future!
You can build your own microscope by using these instructions from Mark Miodownik from Dara O Briain's Science Club on BBC Two. Don't forget to train a model on some microscopic images of thing like corn stem, different fabrics, or blood. Use our training platform and add images of whatever you'd like.
Download Project Launch Codelab
Project Guide
-
Step 1: This project uses the vision extension and the video extension. The vision extension can detect images based on trained models using our training platform.
-
Step 2: First, download the project. If you've made a webcam microscope and plugged it into a USB port to your laptop, you need to change your video source to the microscope. You can test it out by changing the number on this block until you see your microscope output. This block can be found in the "video-sensing" extension pictured above.
-
Step 3: Here is the first set of blocks that you can build to experiment with your model. Don't forget to put in your api key and the corresponding model for your microscope to use to predict images.
-
Step 4: Here is the second set of blocks needed to predict the images in the microscope.
-
Step 5: After prediction, we need to do something about the results. Add this onto the second set of blocks in step 4. As you can see, you can click the red circle or the green circle on the stage depending on if the prediction was right or wrong.
-
Step 6: If the model was wrong, and you click on the red circle, you are led to this set of blocks belonging to the red circle sprite. By answering the question with the correct category of your picture, you make your model better.
-
Step 7: Here's an example using the live view of a wecbcam microscope.
Disclaimer - the gif above uses a previous version of the Cognimates stack that no longer has the same characters, but the functionality of blocks has remained the same.