Is it a cat? - incorrect Probability

Part 1 2022
1

Hi,
I am a newbie
Just started with the course 3 days back. trying to generate classifier for Cats based on Lesson 1. Everything went well, but last cell says its a cat, with probability 0? Shouldn’t it be 1?

image
image1065×361 17.3 KB

Not sure what I am missing on

4 Likes
2

My guess at what is happening here is 0 (in probs[0]) is not the correct index for the cat class.

Try running this code:

is_cat, idx, probs = learn.predict('cat.jpg')
print(f"This is a: {is_cat}.")
print(f"Probability it's a cat: {probs[idx]:.4f}") # use `idx` instead of 0
5 Likes
3

It worked. Thanks a lot!
image

5 Likes
4

I also tested my model with “cute” cat picture.

image
image1095×645 142 KB

  • is_cat : stores our model’s prediction (about the image) in bool.
    (True: Yes its cat) , (False:its not a cat).

  • _ : stores a binary classification (not cat/ is cat) = ( 0 / 1 )

  • prob : stores an array of probabilities for each class ~ (0/1) ~ (not cat/ is cat)
    i.e. probablity of image not being a cat is very low (0.2282) ,
    and probability of image being a cat is high (0.7718)

Hence ,

  • When its NOT a cat , prob[idx=0] tells what is prob. of image not being a cat.
  • When its a cat, prob[idx=1] tells what is the prob. of image being a cat.
4 Likes