<aside> 💡 [Insturction]

Using the ml5.js examples, try running image classification on a variety of images. Pick at least 10 objects in your room. How many of these does it recognize? What other aspects of the image affect the classification, including but not limited to position, scale, lighting, etc.

</aside>

Result:

I used webcam image classification to detect my daily items, and the model classified four items correctly out of ten. For effective comparison, I tried using the model in the following environment setups:

1. With/Without Background noise
   1. For setting the ‘without background noise,’ I put the camera in front of the wall
2. Dark and bright
3. Close/far distance from the item to the camera

The result told me the model reads items better in the following situations:

1. No background noise:
   1. Background influences the classification because it might distract the model from reading the object.
   2. The model classified the item more accurately when there was no background.
2. Big in scale:
   1. The closer the object is to the camera (big in scale), the more accurate classification the model generates.
3. Bright environment:
   1. The model detected the items more accurately when the environment was bright (optimal daily light).
   2. The image/camera resolution was clear.

Other insights are:

1. Depending on the angle and position, the model might classify the same object in several ways.
2. The more distinctive the item’s features are (i.e., remote controller), the more correct answer is generated.