Diego Marcos

Robert van der Klis, Stephan Alaniz, Massimiliano Mancini et al.

Fine-grained classification often requires recognizing specific object parts, such as beak shape and wing patterns for birds. Encouraging a fine-grained classification model to first detect such parts and then using them to infer the class could help us gauge whether the model is indeed looking at the right details better than with interpretability methods that provide a single attribution map. We propose PDiscoNet to discover object parts by using only image-level class labels along with priors encouraging the parts to be: discriminative, compact, distinct from each other, equivariant to rigid transforms, and active in at least some of the images. In addition to using the appropriate losses to encode these priors, we propose to use part-dropout, where full part feature vectors are dropped at once to prevent a single part from dominating in the classification, and part feature vector modulation, which makes the information coming from each part distinct from the perspective of the classifier. Our results on CUB, CelebA, and PartImageNet show that the proposed method provides substantially better part discovery performance than previous methods while not requiring any additional hyper-parameter tuning and without penalizing the classification performance. The code is available at https://github.com/robertdvdk/part_detection.

Stephan Alaniz, Massimiliano Mancini, Anjan Dutta et al.

Humans show high-level of abstraction capabilities in games that require quickly communicating object information. They decompose the message content into multiple parts and communicate them in an interpretable protocol. Toward equipping machines with such capabilities, we propose the Primitive-based Sketch Abstraction task where the goal is to represent sketches using a fixed set of drawing primitives under the influence of a budget. To solve this task, our Primitive-Matching Network (PMN), learns interpretable abstractions of a sketch in a self supervised manner. Specifically, PMN maps each stroke of a sketch to its most similar primitive in a given set, predicting an affine transformation that aligns the selected primitive to the target stroke. We learn this stroke-to-primitive mapping end-to-end with a distance-transform loss that is minimal when the original sketch is precisely reconstructed with the predicted primitives. Our PMN abstraction empirically achieves the highest performance on sketch recognition and sketch-based image retrieval given a communication budget, while at the same time being highly interpretable. This opens up new possibilities for sketch analysis, such as comparing sketches by extracting the most relevant primitives that define an object category. Code is available at https://github.com/ExplainableML/sketch-primitives.

Hugo Porta, Emanuele Dalsasso, Diego Marcos et al.

Prototypical part learning is emerging as a promising approach for making semantic segmentation interpretable. The model selects real patches seen during training as prototypes and constructs the dense prediction map based on the similarity between parts of the test image and the prototypes. This improves interpretability since the user can inspect the link between the predicted output and the patterns learned by the model in terms of prototypical information. In this paper, we propose a method for interpretable semantic segmentation that leverages multi-scale image representation for prototypical part learning. First, we introduce a prototype layer that explicitly learns diverse prototypical parts at several scales, leading to multi-scale representations in the prototype activation output. Then, we propose a sparse grouping mechanism that produces multi-scale sparse groups of these scale-specific prototypical parts. This provides a deeper understanding of the interactions between multi-scale object representations while enhancing the interpretability of the segmentation model. The experiments conducted on Pascal VOC, Cityscapes, and ADE20K demonstrate that the proposed method increases model sparsity, improves interpretability over existing prototype-based methods, and narrows the performance gap with the non-interpretable counterpart models. Code is available at github.com/eceo-epfl/ScaleProtoSeg.

Konstantinos P. Panousis, Dino Ienco, Diego Marcos

The recent mass adoption of DNNs, even in safety-critical scenarios, has shifted the focus of the research community towards the creation of inherently intrepretable models. Concept Bottleneck Models (CBMs) constitute a popular approach where hidden layers are tied to human understandable concepts allowing for investigation and correction of the network's decisions. However, CBMs usually suffer from: (i) performance degradation and (ii) lower interpretability than intended due to the sheer amount of concepts contributing to each decision. In this work, we propose a simple yet highly intuitive interpretable framework based on Contrastive Language Image models and a single sparse linear layer. In stark contrast to related approaches, the sparsity in our framework is achieved via principled Bayesian arguments by inferring concept presence via a data-driven Bernoulli distribution. As we experimentally show, our framework not only outperforms recent CBM approaches accuracy-wise, but it also yields high per example concept sparsity, facilitating the individual investigation of the emerging concepts.

Diego Marcos, Robert van de Vlasakker, Ioannis N. Athanasiadis et al.

Plant morphological traits, their observable characteristics, are fundamental to understand the role played by each species within their ecosystem. However, compiling trait information for even a moderate number of species is a demanding task that may take experts years to accomplish. At the same time, massive amounts of information about species descriptions is available online in the form of text, although the lack of structure makes this source of data impossible to use at scale. To overcome this, we propose to leverage recent advances in large language models (LLMs) and devise a mechanism for gathering and processing information on plant traits in the form of unstructured textual descriptions, without manual curation. We evaluate our approach by automatically replicating three manually created species-trait matrices. Our method managed to find values for over half of all species-trait pairs, with an F1-score of over 75%. Our results suggest that large-scale creation of structured trait databases from unstructured online text is currently feasible thanks to the information extraction capabilities of LLMs, being limited by the availability of textual descriptions covering all the traits of interest.

Dilli R. Paudel, Diego Marcos, Allard de Wit et al.

Predictor inputs and label data for crop yield forecasting are not always available at the same spatial resolution. We propose a deep learning framework that uses high resolution inputs and low resolution labels to produce crop yield forecasts for both spatial levels. The forecasting model is calibrated by weak supervision from low resolution crop area and yield statistics. We evaluated the framework by disaggregating regional yields in Europe from parent statistical regions to sub-regions for five countries (Germany, Spain, France, Hungary, Italy) and two crops (soft wheat and potatoes). Performance of weakly supervised models was compared with linear trend models and Gradient-Boosted Decision Trees (GBDT). Higher resolution crop yield forecasts are useful to policymakers and other stakeholders. Weakly supervised deep learning methods provide a way to produce such forecasts even in the absence of high resolution yield data.

Ananthu Aniraj, Cassio F. Dantas, Dino Ienco et al.

Models for fine-grained image classification tasks, where the difference between some classes can be extremely subtle and the number of samples per class tends to be low, are particularly prone to picking up background-related biases and demand robust methods to handle potential examples with out-of-distribution (OOD) backgrounds. To gain deeper insights into this critical problem, our research investigates the impact of background-induced bias on fine-grained image classification, evaluating standard backbone models such as Convolutional Neural Network (CNN) and Vision Transformers (ViT). We explore two masking strategies to mitigate background-induced bias: Early masking, which removes background information at the (input) image level, and late masking, which selectively masks high-level spatial features corresponding to the background. Extensive experiments assess the behavior of CNN and ViT models under different masking strategies, with a focus on their generalization to OOD backgrounds. The obtained findings demonstrate that both proposed strategies enhance OOD performance compared to the baseline models, with early masking consistently exhibiting the best OOD performance. Notably, a ViT variant employing GAP-Pooled Patch token-based classification combined with early masking achieves the highest OOD robustness.

Lukas Picek, Christophe Botella, Maximilien Servajean et al.

The difficulty of monitoring biodiversity at fine scales and over large areas limits ecological knowledge and conservation efforts. To fill this gap, Species Distribution Models (SDMs) predict species across space from spatially explicit features. Yet, they face the challenge of integrating the rich but heterogeneous data made available over the past decade, notably millions of opportunistic species observations and standardized surveys, as well as multimodal remote sensing data. In light of that, we have designed and developed a new European-scale dataset for SDMs at high spatial resolution (10--50m), including more than 10k species (i.e., most of the European flora). The dataset comprises 5M heterogeneous Presence-Only records and 90k exhaustive Presence-Absence survey records, all accompanied by diverse environmental rasters (e.g., elevation, human footprint, and soil) traditionally used in SDMs. In addition, it provides Sentinel-2 RGB and NIR satellite images with 10 m resolution, a 20-year time series of climatic variables, and satellite time series from the Landsat program. In addition to the data, we provide an openly accessible SDM benchmark (hosted on Kaggle), which has already attracted an active community and a set of strong baselines for single predictor/modality and multimodal approaches. All resources, e.g., the dataset, pre-trained models, and baseline methods (in the form of notebooks), are available on Kaggle, allowing one to start with our dataset literally with two mouse clicks.

Nicolas Houdré, Diego Marcos, Hugo Riffaud de Turckheim et al.

Earth observation (EO) data spans a wide range of spatial, spectral, and temporal resolutions, from high-resolution optical imagery to low resolution multispectral products or radar time series. While recent foundation models have improved multimodal integration for learning meaningful representations, they often expect fixed input resolutions or are based on sensor-specific encoders limiting generalization across heterogeneous EO modalities. To overcome these limitations we introduce RAMEN, a resolution-adjustable multimodal encoder that learns a shared visual representation across EO data in a fully sensor-agnostic manner. RAMEN treats the modality and spatial and temporal resolutions as key input data features, enabling coherent analysis across modalities within a unified latent space. Its main methodological contribution is to define spatial resolution as a controllable output parameter, giving users direct control over the desired level of detail at inference and allowing explicit trade-offs between spatial precision and computational cost. We train a single, unified transformer encoder reconstructing masked multimodal EO data drawn from diverse sources, ensuring generalization across sensors and resolutions. Once pretrained, RAMEN transfers effectively to both known and unseen sensor configurations and outperforms larger state-of-the-art models on the community-standard PANGAEA benchmark, containing various multi-sensor and multi-resolution downstream tasks. Our code and pretrained model are available at https://github.com/nicolashoudre/RAMEN.

Cassio F. Dantas, Diego Marcos, Dino Ienco

Counterfactual explanations are an emerging tool to enhance interpretability of deep learning models. Given a sample, these methods seek to find and display to the user similar samples across the decision boundary. In this paper, we propose a generative adversarial counterfactual approach for satellite image time series in a multi-class setting for the land cover classification task. One of the distinctive features of the proposed approach is the lack of prior assumption on the targeted class for a given counterfactual explanation. This inherent flexibility allows for the discovery of interesting information on the relationship between land cover classes. The other feature consists of encouraging the counterfactual to differ from the original sample only in a small and compact temporal segment. These time-contiguous perturbations allow for a much sparser and, thus, interpretable solution. Furthermore, plausibility/realism of the generated counterfactual explanations is enforced via the proposed adversarial learning strategy.

Konstantinos P. Panousis, Dino Ienco, Diego Marcos

Deep learning algorithms have recently gained significant attention due to their impressive performance. However, their high complexity and un-interpretable mode of operation hinders their confident deployment in real-world safety-critical tasks. This work targets ante hoc interpretability, and specifically Concept Bottleneck Models (CBMs). Our goal is to design a framework that admits a highly interpretable decision making process with respect to human understandable concepts, on two levels of granularity. To this end, we propose a novel two-level concept discovery formulation leveraging: (i) recent advances in vision-language models, and (ii) an innovative formulation for coarse-to-fine concept selection via data-driven and sparsity-inducing Bayesian arguments. Within this framework, concept information does not solely rely on the similarity between the whole image and general unstructured concepts; instead, we introduce the notion of concept hierarchy to uncover and exploit more granular concept information residing in patch-specific regions of the image scene. As we experimentally show, the proposed construction not only outperforms recent CBM approaches, but also yields a principled framework towards interpetability.

Valerie Zermatten, Chiara Vanalli, Gencer Sumbul et al.

Recent developments in natural language processing highlight text as an emerging data source for ecology. Textual resources carry unique information that can be used in complementarity with geospatial data sources, thus providing insights at the local scale into environmental conditions and properties hidden from more traditional data sources. Leveraging textual information in a spatial context presents several challenges. First, the contribution of textual data remains poorly defined in an ecological context, and it is unclear for which tasks it should be incorporated. Unlike ubiquitous satellite imagery or environmental covariates, the availability of textual data is sparse and irregular; its integration with geospatial data is not straightforward. In response to these challenges, this work proposes an attention-based approach that combines aerial imagery and geolocated text within a spatial neighbourhood, i.e. integrating contributions from several nearby observations. Our approach combines vision and text representations with a geolocation encoding, with an attention-based module that dynamically selects spatial neighbours that are useful for predictive tasks.The proposed approach is applied to the EcoWikiRS dataset, which combines high-resolution aerial imagery with sentences extracted from Wikipedia describing local environmental conditions across Switzerland. Our model is evaluated on the task of predicting 103 environmental variables from the SWECO25 data cube. Our approach consistently outperforms single-location or unimodal, i.e. image-only or text-only, baselines. When analysing variables by thematic groups, results show a significant improvement in performance for climatic, edaphic, population and land use/land cover variables, underscoring the benefit of including the spatial context when combining text and image data.

Ananthu Aniraj, Cassio F. Dantas, Dino Ienco et al.

Computer vision methods that explicitly detect object parts and reason on them are a step towards inherently interpretable models. Existing approaches that perform part discovery driven by a fine-grained classification task make very restrictive assumptions on the geometric properties of the discovered parts; they should be small and compact. Although this prior is useful in some cases, in this paper we show that pre-trained transformer-based vision models, such as self-supervised DINOv2 ViT, enable the relaxation of these constraints. In particular, we find that a total variation (TV) prior, which allows for multiple connected components of any size, substantially outperforms previous work. We test our approach on three fine-grained classification benchmarks: CUB, PartImageNet and Oxford Flowers, and compare our results to previously published methods as well as a re-implementation of the state-of-the-art method PDiscoNet with a transformer-based backbone. We consistently obtain substantial improvements across the board, both on part discovery metrics and the downstream classification task, showing that the strong inductive biases in self-supervised ViT models require to rethink the geometric priors that can be used for unsupervised part discovery.

Valerie Zermatten, Javiera Castillo-Navarro, Pallavi Jain et al.

The presence of species provides key insights into the ecological properties of a location such as land cover, climatic conditions or even soil properties. We propose a method to predict such ecological properties directly from remote sensing (RS) images by aligning them with species habitat descriptions. We introduce the EcoWikiRS dataset, consisting of high-resolution aerial images, the corresponding geolocated species observations, and, for each species, the textual descriptions of their habitat from Wikipedia. EcoWikiRS offers a scalable way of supervision for RS vision language models (RS-VLMs) for ecology. This is a setting with weak and noisy supervision, where, for instance, some text may describe properties that are specific only to part of the species' niche or is irrelevant to a specific image. We tackle this by proposing WINCEL, a weighted version of the InfoNCE loss. We evaluate our model on the task of ecosystem zero-shot classification by following the habitat definitions from the European Nature Information System (EUNIS). Our results show that our approach helps in understanding RS images in a more ecologically meaningful manner. The code and the dataset are available at https://github.com/eceo-epfl/EcoWikiRS.

Konstantinos P. Panousis, Diego Marcos

The widespread adoption of Vision-Language Models (VLMs) across fields has amplified concerns about model interpretability. Distressingly, these models are often treated as black-boxes, with limited or non-existent investigation of their decision making process. Despite numerous post- and ante-hoc interepretability methods, systematic and objective evaluation of the learned representations remains limited, particularly for sparsity-aware methods that are increasingly considered to "induce interpretability". In this work, we focus on Concept Bottleneck Models and investigate how different modeling decisions affect the emerging representations. We introduce the notion of clarity, a measure, capturing the interplay between the downstream performance and the sparsity and precision of the concept representation, while proposing an interpretability assessment framework using datasets with ground truth concept annotations. We consider both VLM- and attribute predictor-based CBMs, and three different sparsity-inducing strategies: per example $\ell_1, \ell_0$ and Bernoulli-based formulations. Our experiments reveal a critical trade-off between flexibility and interpretability, under which a given method can exhibit markedly different behaviors even at comparable performance levels. The code will be made publicly available upon publication.

Ananthu Aniraj, Cassio F. Dantas, Dino Ienco et al.

We introduce an attention-based method that uses learned binary attention masks to ensure that only attended image regions influence the prediction. Context can strongly affect object perception, sometimes leading to biased representations, particularly when objects appear in out-of-distribution backgrounds. At the same time, many image-level object-centric tasks require identifying relevant regions, often requiring context. To address this conundrum, we propose a two-stage framework: stage 1 processes the full image to discover object parts and identify task-relevant regions, while stage 2 leverages input attention masking to restrict its receptive field to these regions, enabling a focused analysis while filtering out potentially spurious information. Both stages are trained jointly, allowing stage 2 to refine stage 1. Extensive experiments across diverse benchmarks demonstrate that our approach significantly improves robustness against spurious correlations and out-of-distribution backgrounds. Code: https://github.com/ananthu-aniraj/ifam

Pallavi Jain, Diego Marcos, Dino Ienco et al.

Vision-language models have shown significant promise in remote sensing applications, particularly for land-use and land-cover (LULC) via zero-shot classification and retrieval. However, current approaches face two key challenges: reliance on large spatial tiles that increase computational cost, and dependence on text-based supervision, which is often not readily available. In this work, we present TimeSenCLIP, a lightweight framework that reevaluate the role of spatial context by evaluating the effectiveness of a single pixel by leveraging its temporal and spectral dimensions, for classifying LULC and ecosystem types. By leveraging spectral and temporal information from Sentinel-2 imagery and cross-view learning with geo-tagged ground-level photos, we minimises the need for caption-based training while preserving semantic alignment between overhead (satellite) and ground perspectives. Our approach is grounded in the LUCAS and Sen4Map datasets, and evaluated on classification tasks including LULC, crop type, and ecosystem type. We demonstrate that single pixel inputs, when combined with temporal and spectral cues, are sufficient for thematic mapping, offering a scalable and efficient alternative for large-scale remote sensing applications. Code is available at https://github.com/pallavijain-pj/TimeSenCLIP

Ananthu Aniraj, Cassio F. Dantas, Dino Ienco et al.

Part-based reasoning is a classical strategy to make a computer vision model directly focus on the object parts that are relevant to the downstream task. In the context of deep learning, this also serves to improve by-design interpretability, often by using part-centric attention mechanisms on top of a latent image representation provided by a standard, black-box model. This approach is based on a locality assumption: that the latent representation of an object part encodes primarily information about the corresponding image region. In this work, we test this basic assumption, measuring intra-object leakage in vision models using part-based attribute annotations. Through a comprehensive experimental evaluation, we show that modern pretrained vision transformers violate the locality assumption and exhibit a strong intra-object leakage, in which each part encodes information from the whole object, a visual metonymy that compromises the faithfulness of attention-based interpretable-by-design methods for part-based reasoning, ultimately rendering them uninterpretable. In addition, we establish an upper bound using a two-stage approach that prevents leakage by design. We then show that this inherently disentangled feature extraction improves attribute-driven part discovery on a variety of tasks, confirming the practical impact of intra-object leakage. Our results uncover a neglected issue affecting the interpretability of part-based representations, such as those in CBMs relying on part-centric concepts, highlighting that two-stage approaches offer a promising way to mitigate it.

Christophe Botella, Benjamin Deneu, Diego Marcos et al.

Understanding the spatio-temporal distribution of species is a cornerstone of ecology and conservation. By pairing species observations with geographic and environmental predictors, researchers can model the relationship between an environment and the species which may be found there. To advance the state-of-the-art in this area with deep learning models and remote sensing data, we organized an open machine learning challenge called GeoLifeCLEF 2023. The training dataset comprised 5 million plant species observations (single positive label per sample) distributed across Europe and covering most of its flora, high-resolution rasters: remote sensing imagery, land cover, elevation, in addition to coarse-resolution data: climate, soil and human footprint variables. In this multi-label classification task, we evaluated models ability to predict the species composition in 22 thousand small plots based on standardized surveys. This paper presents an overview of the competition, synthesizes the approaches used by the participating teams, and analyzes the main results. In particular, we highlight the biases faced by the methods fitted to single positive labels when it comes to the multi-label evaluation, and the new and effective learning strategy combining single and multi-label data in training.

Pallavi Jain, Dino Ienco, Roberto Interdonato et al.

Pre-trained vision-language models (VLMs), such as CLIP, demonstrate impressive zero-shot classification capabilities with free-form prompts and even show some generalization in specialized domains. However, their performance on satellite imagery is limited due to the underrepresentation of such data in their training sets, which predominantly consist of ground-level images. Existing prompting techniques for satellite imagery are often restricted to generic phrases like a satellite image of ..., limiting their effectiveness for zero-shot land-use and land-cover (LULC) mapping. To address these challenges, we introduce SenCLIP, which transfers CLIPs representation to Sentinel-2 imagery by leveraging a large dataset of Sentinel-2 images paired with geotagged ground-level photos from across Europe. We evaluate SenCLIP alongside other SOTA remote sensing VLMs on zero-shot LULC mapping tasks using the EuroSAT and BigEarthNet datasets with both aerial and ground-level prompting styles. Our approach, which aligns ground-level representations with satellite imagery, demonstrates significant improvements in classification accuracy across both prompt styles, opening new possibilities for applying free-form textual descriptions in zero-shot LULC mapping.

Ron van Bree, Diego Marcos, Ioannis Athanasiadis

Biophysical models offer valuable insights into climate-phenology relationships in both natural and agricultural settings. However, there are substantial structural discrepancies across models which require site-specific recalibration, often yielding inconsistent predictions under similar climate scenarios. Machine learning methods offer data-driven solutions, but often lack interpretability and alignment with existing knowledge. We present a phenology model describing dormancy in fruit trees, integrating conventional biophysical models with a neural network to address their structural disparities. We evaluate our hybrid model in an extensive case study predicting cherry tree phenology in Japan, South Korea and Switzerland. Our approach consistently outperforms both traditional biophysical and machine learning models in predicting blooming dates across years. Additionally, the neural network's adaptability facilitates parameter learning for specific tree varieties, enabling robust generalization to new sites without site-specific recalibration. This hybrid model leverages both biophysical constraints and data-driven flexibility, offering a promising avenue for accurate and interpretable phenology modeling.

Alex Levering, Diego Marcos, Devis Tuia

In our research we test data and models for the recognition of housing quality in the city of Amsterdam from ground-level and aerial imagery. For ground-level images we compare Google StreetView (GSV) to Flickr images. Our results show that GSV predicts the most accurate building quality scores, approximately 30% better than using only aerial images. However, we find that through careful filtering and by using the right pre-trained model, Flickr image features combined with aerial image features are able to halve the performance gap to GSV features from 30% to 15%. Our results indicate that there are viable alternatives to GSV for liveability factor prediction, which is encouraging as GSV images are more difficult to acquire and not always available.

Hugo Riffaud de Turckheim, Sylvain Lobry, Roberto Interdonato et al.

The growing number of Earth observation satellites has led to increasingly diverse remote sensing data, with varying spatial, spectral, and temporal configurations. Most existing models rely on fixed input formats and modality-specific encoders, which require retraining when new configurations are introduced, limiting their ability to generalize across modalities. We introduce Atomizer, a flexible architecture that represents remote sensing images as sets of scalars, each corresponding to a spectral band value of a pixel. Each scalar is enriched with contextual metadata (acquisition time, spatial resolution, wavelength, and bandwidth), producing an atomic representation that allows a single encoder to process arbitrary modalities without interpolation or resampling. Atomizer uses structured tokenization with Fourier features and non-uniform radial basis functions to encode content and context, and maps tokens into a latent space via cross-attention. Under modality-disjoint evaluations, Atomizer outperforms standard models and demonstrates robust performance across varying resolutions and spatial sizes.

Maxime Ryckewaert, Diego Marcos, Christophe Botella et al.

The rapid expansion of citizen science initiatives has led to a significant growth of biodiversity databases, and particularly presence-only (PO) observations. PO data are invaluable for understanding species distributions and their dynamics, but their use in a Species Distribution Model (SDM) is curtailed by sampling biases and the lack of information on absences. Poisson point processes are widely used for SDMs, with Maxent being one of the most popular methods. Maxent maximises the entropy of a probability distribution across sites as a function of predefined transformations of variables, called features. In contrast, neural networks and deep learning have emerged as a promising technique for automatic feature extraction from complex input variables. Arbitrarily complex transformations of input variables can be learned from the data efficiently through backpropagation and stochastic gradient descent (SGD). In this paper, we propose DeepMaxent, which harnesses neural networks to automatically learn shared features among species, using the maximum entropy principle. To do so, it employs a normalised Poisson loss where for each species, presence probabilities across sites are modelled by a neural network. We evaluate DeepMaxent on a benchmark dataset known for its spatial sampling biases, using PO data for calibration and presence-absence (PA) data for validation across six regions with different biological groups and covariates. Our results indicate that DeepMaxent performs better than Maxent and other leading SDMs across all regions and taxonomic groups. The method performs particularly well in regions of uneven sampling, demonstrating substantial potential to increase SDM performances. In particular, our approach yields more accurate predictions than traditional single-species models, which opens up new possibilities for methodological enhancement.

Ivica Obadic, Alex Levering, Lars Pennig et al.

Predicting socioeconomic indicators from satellite imagery with deep learning has become an increasingly popular research direction. Post-hoc concept-based explanations can be an important step towards broader adoption of these models in policy-making as they enable the interpretation of socioeconomic outcomes based on visual concepts that are intuitive to humans. In this paper, we study the interplay between representation learning using an additional task-specific contrastive loss and post-hoc concept explainability for socioeconomic studies. Our results on two different geographical locations and tasks indicate that the task-specific pretraining imposes a continuous ordering of the latent space embeddings according to the socioeconomic outcomes. This improves the model's interpretability as it enables the latent space of the model to associate concepts encoding typical urban and natural area patterns with continuous intervals of socioeconomic outcomes. Further, we illustrate how analyzing the model's conceptual sensitivity for the intervals of socioeconomic outcomes can shed light on new insights for urban studies.

Alex Levering, Diego Marcos, Devis Tuia

In this paper we explore deep learning models to monitor longitudinal liveability changes in Dutch cities at the neighbourhood level. Our liveability reference data is defined by a country-wise yearly survey based on a set of indicators combined into a liveability score, the Leefbaarometer. We pair this reference data with yearly-available high-resolution aerial images, which creates yearly timesteps at which liveability can be monitored. We deploy a convolutional neural network trained on an aerial image from 2016 and the Leefbaarometer score to predict liveability at new timesteps 2012 and 2020. The results in a city used for training (Amsterdam) and one never seen during training (Eindhoven) show some trends which are difficult to interpret, especially in light of the differences in image acquisitions at the different time steps. This demonstrates the complexity of liveability monitoring across time periods and the necessity for more sophisticated methods compensating for changes unrelated to liveability dynamics.

Marrit Leenstra, Diego Marcos, Francesca Bovolo et al.

While annotated images for change detection using satellite imagery are scarce and costly to obtain, there is a wealth of unlabeled images being generated every day. In order to leverage these data to learn an image representation more adequate for change detection, we explore methods that exploit the temporal consistency of Sentinel-2 times series to obtain a usable self-supervised learning signal. For this, we build and make publicly available (https://zenodo.org/record/4280482) the Sentinel-2 Multitemporal Cities Pairs (S2MTCP) dataset, containing multitemporal image pairs from 1520 urban areas worldwide. We test the results of multiple self-supervised learning methods for pre-training models for change detection and apply it on a public change detection dataset made of Sentinel-2 image pairs (OSCD).

Yuansheng Hua, Diego Marcos, Lichao Mou et al.

Training Convolutional Neural Networks (CNNs) for very high resolution images requires a large quantity of high-quality pixel-level annotations, which is extremely labor- and time-consuming to produce. Moreover, professional photo interpreters might have to be involved for guaranteeing the correctness of annotations. To alleviate such a burden, we propose a framework for semantic segmentation of aerial images based on incomplete annotations, where annotators are asked to label a few pixels with easy-to-draw scribbles. To exploit these sparse scribbled annotations, we propose the FEature and Spatial relaTional regulArization (FESTA) method to complement the supervised task with an unsupervised learning signal that accounts for neighbourhood structures both in spatial and feature terms.

Devis Tuia, Diego Marcos, Gustau Camps-Valls

Remote sensing image classification exploiting multiple sensors is a very challenging problem: data from different modalities are affected by spectral distortions and mis-alignments of all kinds, and this hampers re-using models built for one image to be used successfully in other scenes. In order to adapt and transfer models across image acquisitions, one must be able to cope with datasets that are not co-registered, acquired under different illumination and atmospheric conditions, by different sensors, and with scarce ground references. Traditionally, methods based on histogram matching have been used. However, they fail when densities have very different shapes or when there is no corresponding band to be matched between the images. An alternative builds upon \emph{manifold alignment}. Manifold alignment performs a multidimensional relative normalization of the data prior to product generation that can cope with data of different dimensionality (e.g. different number of bands) and possibly unpaired examples. Aligning data distributions is an appealing strategy, since it allows to provide data spaces that are more similar to each other, regardless of the subsequent use of the transformed data. In this paper, we study a methodology that aligns data from different domains in a nonlinear way through {\em kernelization}. We introduce the Kernel Manifold Alignment (KEMA) method, which provides a flexible and discriminative projection map, exploits only a few labeled samples (or semantic ties) in each domain, and reduces to solving a generalized eigenvalue problem. We successfully test KEMA in multi-temporal and multi-source very high resolution classification tasks, as well as on the task of making a model invariant to shadowing for hyperspectral imaging.

Diego Marcos, Ruth Fong, Sylvain Lobry et al.

Convolutional neural networks (CNN) are known to learn an image representation that captures concepts relevant to the task, but do so in an implicit way that hampers model interpretability. However, one could argue that such a representation is hidden in the neurons and can be made explicit by teaching the model to recognize semantically interpretable attributes that are present in the scene. We call such an intermediate layer a \emph{semantic bottleneck}. Once the attributes are learned, they can be re-combined to reach the final decision and provide both an accurate prediction and an explicit reasoning behind the CNN decision. In this paper, we look into semantic bottlenecks that capture context: we want attributes to be in groups of a few meaningful elements and participate jointly to the final decision. We use a two-layer semantic bottleneck that gathers attributes into interpretable, sparse groups, allowing them contribute differently to the final output depending on the context. We test our contextual semantic interpretable bottleneck (CSIB) on the task of landscape scenicness estimation and train the semantic interpretable bottleneck using an auxiliary database (SUN Attributes). Our model yields in predictions as accurate as a non-interpretable baseline when applied to a real-world test set of Flickr images, all while providing clear and interpretable explanations for each prediction.

Sylvain Lobry, Diego Marcos, Jesse Murray et al.

This paper introduces the task of visual question answering for remote sensing data (RSVQA). Remote sensing images contain a wealth of information which can be useful for a wide range of tasks including land cover classification, object counting or detection. However, most of the available methodologies are task-specific, thus inhibiting generic and easy access to the information contained in remote sensing data. As a consequence, accurate remote sensing product generation still requires expert knowledge. With RSVQA, we propose a system to extract information from remote sensing data that is accessible to every user: we use questions formulated in natural language and use them to interact with the images. With the system, images can be queried to obtain high level information specific to the image content or relational dependencies between objects visible in the images. Using an automatic method introduced in this article, we built two datasets (using low and high resolution data) of image/question/answer triplets. The information required to build the questions and answers is queried from OpenStreetMap (OSM). The datasets can be used to train (when using supervised methods) and evaluate models to solve the RSVQA task. We report the results obtained by applying a model based on Convolutional Neural Networks (CNNs) for the visual part and on a Recurrent Neural Network (RNN) for the natural language part to this task. The model is trained on the two datasets, yielding promising results in both cases.

Diego Marcos, Sylvain Lobry, Devis Tuia

A main issue preventing the use of Convolutional Neural Networks (CNN) in end user applications is the low level of transparency in the decision process. Previous work on CNN interpretability has mostly focused either on localizing the regions of the image that contribute to the result or on building an external model that generates plausible explanations. However, the former does not provide any semantic information and the latter does not guarantee the faithfulness of the explanation. We propose an intermediate representation composed of multiple Semantically Interpretable Activation Maps (SIAM) indicating the presence of predefined attributes at different locations of the image. These attribute maps are then linearly combined to produce the final output. This gives the user insight into what the model has seen, where, and a final output directly linked to this information in a comprehensive and interpretable way. We test the method on the task of landscape scenicness (aesthetic value) estimation, using an intermediate representation of 33 attributes from the SUN Attributes database. The results confirm that SIAM makes it possible to understand what attributes in the image are contributing to the final score and where they are located. Since it is based on learning from multiple tasks and datasets, SIAM improve the explanability of the prediction without additional annotation efforts or computational overhead at inference time, while keeping good performances on both the final and intermediate tasks.

Benjamin Kellenberger, Diego Marcos, Sylvain Lobry et al.

We present an Active Learning (AL) strategy for re-using a deep Convolutional Neural Network (CNN)-based object detector on a new dataset. This is of particular interest for wildlife conservation: given a set of images acquired with an Unmanned Aerial Vehicle (UAV) and manually labeled gound truth, our goal is to train an animal detector that can be re-used for repeated acquisitions, e.g. in follow-up years. Domain shifts between datasets typically prevent such a direct model application. We thus propose to bridge this gap using AL and introduce a new criterion called Transfer Sampling (TS). TS uses Optimal Transport to find corresponding regions between the source and the target datasets in the space of CNN activations. The CNN scores in the source dataset are used to rank the samples according to their likelihood of being animals, and this ranking is transferred to the target dataset. Unlike conventional AL criteria that exploit model uncertainty, TS focuses on very confident samples, thus allowing a quick retrieval of true positives in the target dataset, where positives are typically extremely rare and difficult to find by visual inspection. We extend TS with a new window cropping strategy that further accelerates sample retrieval. Our experiments show that with both strategies combined, less than half a percent of oracle-provided labels are enough to find almost 80% of the animals in challenging sets of UAV images, beating all baselines by a margin.

Stephan Alaniz, Diego Marcos, Bernt Schiele et al.

Integrated interpretability without sacrificing the prediction accuracy of decision making algorithms has the potential of greatly improving their value to the user. Instead of assigning a label to an image directly, we propose to learn iterative binary sub-decisions, inducing sparsity and transparency in the decision making process. The key aspect of our model is its ability to build a decision tree whose structure is encoded into the memory representation of a Recurrent Neural Network jointly learned by two models communicating through message passing. In addition, our model assigns a semantic meaning to each decision in the form of binary attributes, providing concise, semantic and relevant rationalizations to the user. On three benchmark image classification datasets, including the large-scale ImageNet, our model generates human interpretable binary decision sequences explaining the predictions of the network while maintaining state-of-the-art accuracy.

Diego Marcos, Benjamin Kellenberger, Sylvain Lobry et al.

We study the effect of injecting local scale equivariance into Convolutional Neural Networks. This is done by applying each convolutional filter at multiple scales. The output is a vector field encoding for the maximally activating scale and the scale itself, which is further processed by the following convolutional layers. This allows all the intermediate representations to be locally scale equivariant. We show that this improves the performance of the model by over $20\%$ in the scale equivariant task of regressing the scaling factor applied to randomly scaled MNIST digits. Furthermore, we find it also useful for scale invariant tasks, such as the actual classification of randomly scaled digits. This highlights the usefulness of allowing for a compact representation that can also learn relationships between different local scales by keeping internal scale equivariance.

Benjamin Kellenberger, Diego Marcos, Devis Tuia

Knowledge over the number of animals in large wildlife reserves is a vital necessity for park rangers in their efforts to protect endangered species. Manual animal censuses are dangerous and expensive, hence Unmanned Aerial Vehicles (UAVs) with consumer level digital cameras are becoming a popular alternative tool to estimate livestock. Several works have been proposed that semi-automatically process UAV images to detect animals, of which some employ Convolutional Neural Networks (CNNs), a recent family of deep learning algorithms that proved very effective in object detection in large datasets from computer vision. However, the majority of works related to wildlife focuses only on small datasets (typically subsets of UAV campaigns), which might be detrimental when presented with the sheer scale of real study areas for large mammal census. Methods may yield thousands of false alarms in such cases. In this paper, we study how to scale CNNs to large wildlife census tasks and present a number of recommendations to train a CNN on a large UAV dataset. We further introduce novel evaluation protocols that are tailored to censuses and model suitability for subsequent human verification of detections. Using our recommendations, we are able to train a CNN reducing the number of false positives by an order of magnitude compared to previous state-of-the-art. Setting the requirements at 90% recall, our CNN allows to reduce the amount of data required for manual verification by three times, thus making it possible for rangers to screen all the data acquired efficiently and to detect almost all animals in the reserve automatically.

Diego Marcos, Devis Tuia, Benjamin Kellenberger et al.

The world is covered with millions of buildings, and precisely knowing each instance's position and extents is vital to a multitude of applications. Recently, automated building footprint segmentation models have shown superior detection accuracy thanks to the usage of Convolutional Neural Networks (CNN). However, even the latest evolutions struggle to precisely delineating borders, which often leads to geometric distortions and inadvertent fusion of adjacent building instances. We propose to overcome this issue by exploiting the distinct geometric properties of buildings. To this end, we present Deep Structured Active Contours (DSAC), a novel framework that integrates priors and constraints into the segmentation process, such as continuous boundaries, smooth edges, and sharp corners. To do so, DSAC employs Active Contour Models (ACM), a family of constraint- and prior-based polygonal models. We learn ACM parameterizations per instance using a CNN, and show how to incorporate all components in a structured output model, making DSAC trainable end-to-end. We evaluate DSAC on three challenging building instance segmentation datasets, where it compares favorably against state-of-the-art. Code will be made available.

Diego Marcos, Michele Volpi, Benjamin Kellenberger et al.

In remote sensing images, the absolute orientation of objects is arbitrary. Depending on an object's orientation and on a sensor's flight path, objects of the same semantic class can be observed in different orientations in the same image. Equivariance to rotation, in this context understood as responding with a rotated semantic label map when subject to a rotation of the input image, is therefore a very desirable feature, in particular for high capacity models, such as Convolutional Neural Networks (CNNs). If rotation equivariance is encoded in the network, the model is confronted with a simpler task and does not need to learn specific (and redundant) weights to address rotated versions of the same object class. In this work we propose a CNN architecture called Rotation Equivariant Vector Field Network (RotEqNet) to encode rotation equivariance in the network itself. By using rotating convolutions as building blocks and passing only the the values corresponding to the maximally activating orientation throughout the network in the form of orientation encoding vector fields, RotEqNet treats rotated versions of the same object with the same filter bank and therefore achieves state-of-the-art performances even when using very small architectures trained from scratch. We test RotEqNet in two challenging sub-decimeter resolution semantic labeling problems, and show that we can perform better than a standard CNN while requiring one order of magnitude less parameters.

Diego Marcos, Michele Volpi, Nikos Komodakis et al.

In many computer vision tasks, we expect a particular behavior of the output with respect to rotations of the input image. If this relationship is explicitly encoded, instead of treated as any other variation, the complexity of the problem is decreased, leading to a reduction in the size of the required model. In this paper, we propose the Rotation Equivariant Vector Field Networks (RotEqNet), a Convolutional Neural Network (CNN) architecture encoding rotation equivariance, invariance and covariance. Each convolutional filter is applied at multiple orientations and returns a vector field representing magnitude and angle of the highest scoring orientation at every spatial location. We develop a modified convolution operator relying on this representation to obtain deep architectures. We test RotEqNet on several problems requiring different responses with respect to the inputs' rotation: image classification, biomedical image segmentation, orientation estimation and patch matching. In all cases, we show that RotEqNet offers extremely compact models in terms of number of parameters and provides results in line to those of networks orders of magnitude larger.

Diego Marcos, Michele Volpi, Devis Tuia

We present a method for learning discriminative filters using a shallow Convolutional Neural Network (CNN). We encode rotation invariance directly in the model by tying the weights of groups of filters to several rotated versions of the canonical filter in the group. These filters can be used to extract rotation invariant features well-suited for image classification. We test this learning procedure on a texture classification benchmark, where the orientations of the training images differ from those of the test images. We obtain results comparable to the state-of-the-art. Compared to standard shallow CNNs, the proposed method obtains higher classification performance while reducing by an order of magnitude the number of parameters to be learned.