What is MNIST
The MNIST dataset is a publicly available dataset that serves as the hello world of deep learning, used as a benchmark to evaluate whether a model/library/framework is effective.
The MNIST dataset consists of images of handwritten digits (0-9) and their corresponding labels, comprising 60,000 training samples and 10,000 testing samples. Each sample is a 28 x 28 pixel grayscale image of a handwritten digit. The MNIST dataset is derived from the National Institute of Standards and Technology in the United States, with the entire training set consisting of handwritten digits from 250 different individuals, where 50% are from high school students and 50% are from census workers.
Importing Transforms Method
Import the transforms method and change the transform in the MNIST dataset to transforms.ToTensor():
Partial results of execution:
Combining Transforms:
Partial results of execution:
Conclusion
Transforms are a commonly used image transformation method that can be combined using the Compose method, allowing multiple transforms to process images. The transforms method provides fine-tuned image processing, for example, in segmentation tasks, where you need to build a more complex transformation pipeline, making the transforms method very useful.