Are We Recognizing the Jaguar or Its Background? A Diagnostic Framework for Jaguar Re-Identification

Jaguar re-identification (re-ID) from citizen-science imagery can look strong on standard retrieval metrics while still relying on the wrong evidence, such as background context or silhouette shape, instead of the coat pattern that defines identity. We introduce a diagnostic framework for wildlife re-ID with two axes: a leakage-controlled context ratio, background/foreground, computed from inpainted background-only versus foreground-only images, and a laterality diagnostic based on cross-flank retrieval and mirror self-similarity. To make these diagnostics measurable, we curate a Pantanal jaguar benchmark with per-pixel segmentation masks and an identity-balanced evaluation protocol. We then use representative mitigation families, ArcFace fine-tuning, anti-symmetry regularization, and Lorentz hyperbolic embeddings, as case studies under the same evaluation lens. The goal is not only to ask which model ranks best, but also what visual evidence it uses to do so.

Identification of Seed Cells in Multispectral Images for GrowCut Segmentation

The segmentation of satellite images is a necessary step to perform object-oriented image classification, which has become relevant due to its applicability on images with a high spatial resolution. To perform object-oriented image classification, the studied image must first be segmented in uniform regions. This segmentation requires manual work by an expert user, who must exhaustively explore the image to establish thresholds that generate useful and representative segments without oversegmenting and without discarding representative segments. We propose a technique that automatically segments the multispectral image while facing these issues. We identify in the image homogenous zones according to their spectral signatures through the use of morphological filters. These homogenous zones are representatives of different types of land coverings in the image and are used as seeds for the GrowCut multispectral segmentation algorithm. GrowCut is a cellular automaton with competitive region growth, its cells are linked to every pixel in the image through three parameters: the pixel's spectral signature, a label, and a strength factor that represents the strength with which a cell defends its label. The seed cells possess maximum strength and maintain their state throughout the automaton's evolution. Starting from seed cells, each cell in the image is iteratively attacked by its neighboring cells. When the automaton stops updating its states, we obtain a segmented image where each pixel has taken the label of one of its cells. In this paper the algorithm was applied in an image acquired by Landsat8 on agricultural land of Calabozo, Guarico, Venezuela where there are different types of land coverings: agriculture, urban regions, water bodies, and savannas with different degrees of human intervention. The segmentation obtained is presented as irregular polygons enclosing geographical objects.

Unsupervised Segmentation of Multispectral Images with Cellular Automata

Multispectral images acquired by satellites are used to study phenomena on the Earth's surface. Unsupervised classification techniques analyze multispectral image content without considering prior knowledge of the observed terrain; this is done using techniques which group pixels that have similar statistics of digital level distribution in the various image channels. In this paper, we propose a methodology for unsupervised classification based on a deterministic cellular automaton. The automaton is initialized in an unsupervised manner by setting seed cells, selected according to two criteria: to be representative of the spatial distribution of the dominant elements in the image, and to take into account the diversity of spectral signatures in the image. The automaton's evolution is based on an attack rule that is applied simultaneously to all its cells. Among the noteworthy advantages of deterministic cellular automata for multispectral processing of satellite imagery is the consideration of topological information in the image via seed positioning, and the ability to modify the scale of the study.