Lecture 2. Image Classifcatioon

tips

• Python + Numpy
• Google Cloud Tutorial

Image Classifacation

A core task in Computer Vision

문제점 : semantic

Challenges :

• Viewpoint variation,
• Illumination,
• Deformation,
• Occlusion,
• Background cluster,
• Intraclass variation

Data Driven Approach

1. Collect a dataset of images and labels
2. Use Machine Learning to train a classifier
3. Evaluate the classifier on new images

1st classifier : Nearest Neighbor

def train(images, labels):
    #machine learning!
    return model

Memorize all data and labels

def predict(model, test_images):
	#Use model to predict labels
    return test_labels

Predict the label of the most similar training image

• train : making a model
• predict : with the model, making a prediction

Example Dataset : CIFAR10

이전에 보면 L1 디스턴스는 비교 이미지, 나머지가 테스트 이미지

Distance Metic to compare images

정말 간단하면서도 바보같은 방법인가!

test image - training immage = pixel-wise absolute value differences

각 픽셀값을 뺀 다음 전부를 더해서 유사도를 측정한다?

prediction은 빠르지만 trainign이 느리다.

O(N), O(1)

K-nearest Neighbors

Instead of copying label from mearest neighbor, take majority vote from K closest points

choosing K, Matrics

L1 ,L2의 특징이 있고, 문제에 따라 다르게 사용하지만

정답은 둘 다 해보고 더 나은 것을 쓴ㄴ 것.

Hyperparameters

• What is the best value of k to use?
• What is the best distance to use?

These

Setting Hyperparameters

Example of 5-hold cross-validation for the value of k.

k-Nearest Neighbor on images never used.

• 왜곡된 이미지
• curse of dimensionality

K-Nearest Neighbors : Summary

In Image Classification we start with a training set of images and labels, and must predict labels on the test set

The K-Nearest Neighborss classifier preicts labels based on nearest traininng examples.

Distance metric and K are hyperparameters

Linear Classifier

• help us whole NN, CNN
• Hard cases for a linear classifier