• API Reference - ImageDatasetDrift

Image Dataset Drift

This notebooks provides an overview for using and understanding the image dataset drift check, used to detect drift in simple image properties between train and test datasets.

Structure:

• What Is Image Dataset Drift?

• Which Image Properties Are Used?

• Loading The Data

• Run The Check

• Define a Condition

What Is Image Dataset Drift?

Drift is simply a change in the distribution of data over time, and it is also one of the top reasons why machine learning model’s performance degrades over time.

Image dataset drift is a drift that occurs in more than one image property at a time, and may even affect the relationships between those properties, which are undetectable by univariate drift methods.

For more information on drift, please visit our Drift Guide.

How Deepchecks Detects Dataset Drift

This check detects multivariate drift by using a domain classifier. Other methods to detect drift include univariate measures which is used in other checks, such as Image Property Drift check.

Using Properties to Detect Image Drift

In computer vision specifically, we can’t measure drift on the images directly, as the individual pixel has little value when estimating drift. Therefore, we calculate drift on different properties of the image, on which we can directly measure drift.

Which Image Properties Are Used?

Property name	What is it
Aspect Ratio	Ratio between height and width of image (height / width)
Area	Area of image in pixels (height * width)
Brightness	Average intensity of image pixels. Color channels have different weights according to RGB-to-Grayscale formula
RMS Contrast	Contrast of image, calculated by standard deviation of pixels
Mean Red Relative Intensity	Mean over all pixels of the red channel, scaled to their relative intensity in comparison to the other channels [r / (r + g + b)].
Mean Green Relative Intensity	Mean over all pixels of the green channel, scaled to their relative intensity in comparison to the other channels [g / (r + g + b)].
Mean Blue Relative Intensity	Mean over all pixels of the blue channel, scaled to their relative intensity in comparison to the other channels [b / (r + g + b)].

Imports

Note

In this example, we use the pytorch version of the coco dataset and model. In order to run this example using tensorflow, please change the import statements to:

from deepchecks.vision.datasets.detection.coco_tensorflow import load_dataset

import numpy as np

from deepchecks.vision.checks import ImageDatasetDrift
from deepchecks.vision.datasets.detection.coco_torch import load_dataset

Loading the data

train_ds = load_dataset(train=True, object_type='VisionData')
test_ds = load_dataset(train=False, object_type='VisionData')

Run the check

without drift

check = ImageDatasetDrift()
result = check.run(train_dataset=train_ds, test_dataset=test_ds)
result

Processing Train Batches:
|     | 0/1 [Time: 00:00]
Processing Train Batches:
|█████| 1/1 [Time: 00:00]
Processing Train Batches:
|█████| 1/1 [Time: 00:00]

Processing Test Batches:
|     | 0/1 [Time: 00:00]

Processing Test Batches:
|█████| 1/1 [Time: 00:00]

Processing Test Batches:
|█████| 1/1 [Time: 00:00]


Computing Check:
|     | 0/1 [Time: 00:00]


Computing Check:
|█████| 1/1 [Time: 00:00]


Computing Check:
|█████| 1/1 [Time: 00:00]

To display the results in an IDE like PyCharm, you can use the following code:

#  result.show_in_window()

The result will be displayed in a new window.

Insert drift

Now, we will define a custom collate function that will insert a drift to the training set.

def add_brightness(img):
    reverse = 255 - img
    addition_of_brightness = (reverse * 0.2).astype(int)
    return img + addition_of_brightness

drifted_train_ds = load_dataset(train=True, object_type='VisionData')

def created_drifted_collate_function(collate_fn):
    def drifted_collate_function(batch):
        data_dict = collate_fn(batch)
        data_dict['images'] = [add_brightness(np.array(img)) for img in data_dict['images']]
        return data_dict
    return drifted_collate_function


drifted_train_ds._batch_loader.collate_fn = created_drifted_collate_function(drifted_train_ds._batch_loader.collate_fn)

Run the check again

check = ImageDatasetDrift()
result = check.run(train_dataset=drifted_train_ds, test_dataset=test_ds)
result

Processing Train Batches:
|     | 0/1 [Time: 00:00]
Processing Train Batches:
|█████| 1/1 [Time: 00:01]
Processing Train Batches:
|█████| 1/1 [Time: 00:01]

Processing Test Batches:
|     | 0/1 [Time: 00:00]

Processing Test Batches:
|█████| 1/1 [Time: 00:00]

Processing Test Batches:
|█████| 1/1 [Time: 00:00]


Computing Check:
|     | 0/1 [Time: 00:00]


Computing Check:
|█████| 1/1 [Time: 00:00]


Computing Check:
|█████| 1/1 [Time: 00:00]

Define a Condition

Now, we will define a condition that the maximum drift score is less than a certain threshold. In this example we will set the threshold at 0.2. In order to demonstrate the condition, we will use again the original (not drifted) train dataset.

check = ImageDatasetDrift().add_condition_drift_score_less_than(0.2)
result = check.run(train_dataset=train_ds, test_dataset=test_ds).show(show_additional_outputs=False)
result

Processing Train Batches:
|     | 0/1 [Time: 00:00]
Processing Train Batches:
|█████| 1/1 [Time: 00:00]
Processing Train Batches:
|█████| 1/1 [Time: 00:00]

Processing Test Batches:
|     | 0/1 [Time: 00:00]

Processing Test Batches:
|█████| 1/1 [Time: 00:00]

Processing Test Batches:
|█████| 1/1 [Time: 00:00]


Computing Check:
|     | 0/1 [Time: 00:00]


Computing Check:
|█████| 1/1 [Time: 00:00]


Computing Check:
|█████| 1/1 [Time: 00:00]

previous

Heatmap Comparison

next

Image Property Drift