9 papers with code • 0 benchmarks • 1 datasets
Visual Instance Search is the task of retrieving from a database of images the ones that contain an instance of a visual query. It is typically much more challenging than finding images from the database that contain objects belonging to the same category as the object in the query. If the visual query is an image of a shoe, visual Instance Search does not try to find images of shoes, which might differ from the query in shape, color or size, but tries to find images of the exact same shoe as the one in the query image. Visual Instance Search challenges image representations as the features extracted from the images must enable such fine-grained recognition despite variations in viewpoints, scale, position, illumination, etc. Whereas holistic image representations, where each image is mapped to a single high-dimensional vector, are sufficient for coarse-grained similarity retrieval, local features are needed for instance retrieval.
These leaderboards are used to track progress in Instance Search
This work proposes a simple instance retrieval pipeline based on encoding the convolutional features of CNN using the bag of words aggregation scheme (BoW).
In this paper, we go beyond this spatial information and propose a local-aware encoding of convolutional features based on semantic information predicted in the target image.
This work explores attention models to weight the contribution of local convolutional representations for the instance search task.
In addition, the proposed enhancement on the network structure also shows superior performance on the instance segmentation task.
Specifically, we propose GlobalTrack, a pure global instance search based tracker that makes no assumption on the temporal consistency of the target's positions and scales.
To address this issue, we propose a confidence-aware active feedback method (CAAF) that is specifically designed for online RF in interactive INS tasks.
Lawyers, for instance, search for appropriate precedents favorable to their clients, while the number of legal precedents is ever-growing.
$k$-nn search in an rpForest is influenced by two factors: 1) the dispersion of points along the random direction and 2) the number of rpTrees in the rpForest.