Recognizing objects

Try this

Classify the cat on a motorcycle, then detect objects in it. What does each model see?
With Find objects, slide the threshold from 5% to 95% on the cafe photo.
Find a picture where the classifier is confident and wrong.
Use the camera and Analyze live. Frames are processed on your device and never uploaded.

Softmax scores and detection thresholdsPython with NumPy

import numpy as np

def softmax(z):
    e = np.exp(z - z.max())
    return e / e.sum()

# On the second picture every label fits much worse (all scores 6 lower),
# yet the probabilities are identical: softmax only compares labels.
good = np.array([8.0, 6.5, 4.0, 3.0])
print("good fit:", np.round(softmax(good), 3))
print("poor fit:", np.round(softmax(good - 6.0), 3))

# A detector's guesses: label and score. The threshold decides which to show.
guesses = [("car", 0.99), ("car", 0.62), ("person", 0.97), ("potted plant", 0.31)]
for t in (0.9, 0.5, 0.2):
    print(f"threshold {t}:", ", ".join(f"{l} {s:.2f}" for l, s in guesses if s >= t))

good fit: [0.801 0.179 0.015 0.005]
poor fit: [0.801 0.179 0.015 0.005]
threshold 0.9: car 0.99, person 0.97
threshold 0.5: car 0.99, car 0.62, person 0.97
threshold 0.2: car 0.99, car 0.62, person 0.97, potted plant 0.31

Sample photos, all CC0 on Wikimedia Commons: Dog by Joselodos; Cat by Vyshnavi medicharla; Cat on a motorcycle by CAPTAIN RAJU; Breakfast by Andy Li; Street by Mucklagh; Cafe by Andy Li; Corridor by Sindugab; Woodblock print by Utagawa Hiroshige (Metropolitan Museum of Art).