Activity 3. Virtual machine for mushrooms recognition
To create this machine, we will use the free-to-use Teachable Machine platform from Google Inc.
The goal is to train a virtual model that can recognise a certain type of mushroom (in this case, “Peach”) just from a photo file or live capture.
1. Familiarise yourself with the instructions in Worksheet 3. Create a virtual machine for mushroom recognition.
2. Complete the tasks in sequence:
- Task 1. Collect a sufficient number of pictures of poisonous mushrooms. For this purpose, we will use the web page https://manatarka.org/otrovni/. There we find a set of poisonous mushrooms whose pictures we could use to train the virtual machine.
- Task 2. Capture a large number of photos of one edible mushroom – in our case it is the Finch. There are about 150 photos. The goal is to have the virtual machine well trained (this is achieved with a large number of raw data – in this case, photos) in recognizing at least one edible mushroom.
- Task 3. Create two classes of data in the Teachable Machine web platform – one class will be “Peach” and the other will be “Poisonous Mushrooms”. We attach the corresponding picture information to each class.
- Task 4. Actual machine learning.
- Task 5. Testing the machine for recognition, by submitting another photo of Patchouli or showing a similar mushroom live, using the web camera of a computer, phone or other capture tool connected to the computer we are working from.
- Task 6. Drawing conclusions.
To download: Worksheet 3. Create a virtual machine for mushroom recognition
Photos of the Fenugreek Mushroom https://drive.google.com/drive/folders/1YGb_DWL1znFh5lH__l7rNQVx7OdQkeJt?usp=sharing
Pictures of poisonous mushrooms A sample source that can be used is the website www.manatarka.org. There are a variety of original photographs of different types of poisonous mushrooms and more. We have the permission of the site owner to use them.
Link to the poisonous mushrooms page – https://manatarka.org/otrovni/
