Look Up a Result by Object ID

Look Up a Result by Object ID#

This notebook shows how to retrieve the inference result for a specific object using Hyrax’s lookup verb.

We’ll use the pre-computed inference results from the Getting Started notebook, where the built-in HyraxCNN model was trained on the CIFAR-10 dataset.

As always, we start by creating a Hyrax instance.

[1]:

from hyrax import Hyrax

h = Hyrax()

Look up a single result with h.lookup()#

h.lookup() returns the model prediction for a given object ID — useful for a quick sanity check on a specific object or similarity search when used with a vector database.

The results_dir argument is optional. If omitted, Hyrax will automatically use the most recent inference output directory.

Note: Object IDs in Hyrax are always strings, even if they look like numbers.

[2]:

example_id = "00326"
result_directory = "./example_results/getting_started_results"

h.lookup(id=example_id, results_dir=result_directory)

[2]:

array([[-0.66786087, -3.1095357 ,  1.1876142 ,  1.2289915 ,  1.4235748 ,
         0.5107451 ,  1.1837559 ,  0.40907362, -0.5396151 , -0.86201495]],
      dtype=float32)

As a simple exercise we can find the predicted class using numpy.

[3]:

import numpy as np

prediction = h.lookup(id=example_id, results_dir=result_directory)
print(f"Predicted class: {np.argmax(prediction)}")

Predicted class: 4