Siamese coding network and pair similarity prediction for near-duplicate image detection

Abstract

Near-duplicate detection in a dataset involves finding the elements that are closest to a new query element according to a given similarity function and proximity threshold. The brute force approach is very computationally intensive as it evaluates the similarity between the queried item and all items in the dataset. The potential application domain is an image sharing website that checks for plagiarism or piracy every time a new image is uploaded. Among the various approaches, near-duplicate detection was effectively addressed by SimPair LSH (Fisichella et al., in Decker, Lhotská, Link, Spies, Wagner (eds) Database and expert systems applications, Springer, 2014). As the name suggests, SimPair LSH uses locality sensitive hashing (LSH) and computes and stores in advance a small set of near-duplicate pairs present in the dataset and uses them to reduce the candidate set returned for a given query using the Triangle inequality. We develop an algorithm that predicts how the candidate set will be reduced. We also develop a new efficient method for near-duplicate image detection using a deep Siamese coding neural network that is able to extract effective features from images useful for building LSH indices. Extensive experiments on two benchmark datasets confirm the effectiveness of our deep Siamese coding network and prediction algorithm.

Real-Time Multi-Vehicle Multi-Camera Tracking with Graph-Based Tracklet Features

“…In these experiments, the backbone is a Siamese network with a 2,048-dimension. As can be seen from Table 3, a threshold of 0.5 yields the best IDF1 of 0.7521, which is in agreement with previous work ( 52 , 53 ).…”

“…Based on an empirical tuning proposed in Fisichella et al ( 52 , 53 ), the distance threshold τ is set to 0.5 in the construction of the graph feature. It is necessary, however, to conduct an ablation study with different thresholds to confirm this decision.…”

“…As a result of an empirical tuning proposed by Fisichella et al ( 52 , 53 ), it was determined that the distance threshold of τ should be set at 0.5. Human judgment confirmed that 0.5 was the boundary between the same and different images: distances greater than 0.5 were perceived as different pairs, and distances less than 0.5 were perceived as identical pairs.…”

NeuralBeds: Neural embeddings for efficient DNA data compression and optimized similarity search

“…The train-test split is based only on the anchor sequence q and does not consider the comparison sequences . We explored two ANN techniques (FCN and CNN) with two loss functions (triplet loss, and ladder loss) to create vector representations of DNA sequences using a Siamese Neural Network (SNN) [40] architecture as shown in Fig. 3 .…”

Few-Shot Learning for Clinical Natural Language Processing Using Siamese Neural Networks: Algorithm Development and Validation Study

“…A Siamese neural network (SNN), sometimes called a twin neural network, is an artificial neural network that uses 2 parallel, weight-sharing machine learning models to compute comparable embeddings. The SNN architecture has shown promising results as an FSL approach in computer vision for similarity detection [ 8 ] and duplicate identification [ 9 ]. Yet, its usage in NLP has been understudied, and, to the best of our knowledge, there have not been any studies investigating SNNs for clinical NLP.…”