Weekend Learning - Image Searching - Papers

Paper #1 - Learning Fine-grained Image Similarity with Deep Ranking
Key Notes

• Extract features like Gabor filters
• HOG

Triplet Architecture

• Query image, positive image, and negative image
• A triplet contains a query image, a positive image, and a negative image
• Positive image is more similar to the query image than the negative image
• Meaningful and discriminative triplets
• A triplet characterizes the relative similarity relationship for the three images.
• The deep ranking model employs a triplet-based hinge loss ranking function to characterize fine-grained image similarity relationships

Person reid for keras
Person re-ID baseline with triplet loss
Image similarity using Triplet Loss 
BUILDING A REVERSE IMAGE SEARCH ENGINE

Paper #2 - Evaluation of Distance Measures for Feature based Image Registration using AlexNet
Key Notes - Distance measures

• Cityblock distance: Measures the path between the pixels based on four connected neighbourhood.
• Euclidean distance: Most commonly used metric to find the difference, calculates the square root of the sum of the absolute differences between two feature points
• Cosine distance: Finds the normalized dot product of the two feature points
• Minkowski distance: Is a generalization of Euclidean Distance
• Correlation distance: The correlation of feature two points, p and q, with k dimensions
• Cosine dissimilarity measure, followed by correlation, consistently gives better matching and registration across images of various deformations

Similarity Measurement in Image Retrieval

Imagesimilarity

from scipy import signal
cor = signal.correlate2d (im1, im2)

Cosine similarity in Python
Image Retrieval (via Autoencoders / Transfer Learning)
artificio: A suite of computer vision deep learning algorithms
Image similarity using Deep CNN and Curriculum Learning
Forensic Similarity for Digital Images
Geometric Image Correspondence Verification by Dense Pixel Matching
A Fast Compression-based Similarity Measure with Applications to Content-based Image Retrieval
Convolutional neural network architecture for geometric matching

Happy Learning!!!