I live in Prince Edward Island, Canada. The red fox is a member of the dog family (Canidae) and is indigenous to Prince Edward Island. Its normal fur color is orange-red but “silver” (black) and cross color phase also occur.
Prince Edward Island has a large urban population of red fox. Mange is a skin disease and caused by a mite. The disease is contagious and can be spread to domestic dogs. There was a outbreak of mange in 2018 that ravaged the population and has not fully recovered.
Foxes were a common sight in my back yard. I have not seen any this summer so decided to test the impulse using Jim (the cat). I have included images of foxes and took pictures of Jim in various positions around the apartment.
Set Up Edge Impulsehttps://studio.edgeimpulse.com
Click on Create a New Project. You now can name the project.
Then select Images:
And finally Select Classify multiple objects
Data Acquisition
The next step is to create impulse.
Now we move to Images and Generate Feature. The model shows images in the training model.
Now we can use Transfer Learning:
The last step is to generate a API key or use the key that's already there. Keys are located in the Dash Board. Copy the key because you will need it for Balena.
GitHub has the deploy button. After you click on it you will see the Deploy to existing fleet screen.
Add the key you saved from Edge Impulse. Change the EI_COLLECT_MODE from 0 to 1.
Click on Add Devices
You can add your WIFI SSID and password. Then save and download the image. I use Etcher now to burn a image to a SD card. Once the image is complete than insert the card to the Raspberry Pi and connect to power. The logs should look similar to the screenshot below.
If you check Edge Impulse than that device is now connected.
Also a camera view is available.
Running the Model on the Raspberry Pi
SummaryI will move forward to finish this project. Despite bad weather conditions in PEI, I think this model can identify Foxes with mange and possibly provide treatment for them. I see that Edge Impulse is a powerful tool for machine learning and to identify issues with wildlife.
Comments