Selected Publications

  • From Google Scholar...

    • 2025

    ArXiv
    Uncertainty Guarantees on Automated Precision Weeding using Conformal Prediction
    Melki, P., Bombrun, L., Diallo Boubacar, Dias, J., & Da Costa, J. P.
    arXiv preprint arXiv:2501.07185, 2025. [Paper]
    After a detailed presentation of the conformal prediction methodology and the development of a precision spraying pipeline based on a ''conformalized'' neural network and well-defined spraying decision rules, the article evaluates this pipeline on two real-world scenarios: one under in-distribution conditions, the other reflecting a near out-of-distribution setting.

    2024

    CVPR
    The Penalized Inverse Probability Measure for Conformal Classification
    Melki, P., Bombrun, L., Diallo Boubacar, Dias, J., & Da Costa, J. P.
    In Proceedings of the Conference on Computer Vision and Pattern Recognition (CVPR), 2024. [Paper]
    Through toy examples and empirical results on the task of crop and weed image classification in agricultural robotics, the current work shows how PIP-based conformal classifiers exhibit precisely the desired behavior in comparison with other nonconformity measures and strike a good balance between informativeness and efficiency.
    ArXiv
    Active learning for efficient annotation in precision agriculture: A use-case on crop-weed semantic segmentation.
    van Marrewijk, B. M., Dandjinou, C., Rustia, D. J. A., Gonzalez, N. F., Diallo Boubacar & Blok, P. M..
    (ICCV 2023) - ArXiv preprint arXiv:2404.02580, 2024. [Paper]
    This study addresses this research gap by conducting a comparative study of three active learning-based acquisition functions: Bayesian Active Learning by Disagreement (BALD), stochastic-based BALD (PowerBALD), and Random. The acquisition functions were tested on two agricultural datasets: Sugarbeet and Corn-Weed, both containing three semantic classes: background, crop and weed.

    2023

    AgriTech
    Impact of the Environmental Conditions on the Robustness of a Weed Detection Model
    Nicolas Franco Gonzalez, Paul Melki, Boubacar Diallo, Hugo Antoine, Estelle Millet.
    AgriTech Day (SIMA), 2023. [Paper]
    The main contribution of this article is the proposal of a methodology for analyzing the impact of this variability on weed detection quality. We focus only on four intrinsic and external factors that contribute to the variability of conditions (which we call meta-variables) : the vegetation development stage (using the BBCH scale), the soil characteristics, namely its color and humidity level, and the condition of the sky at the moment of image acquisition (sunny or overcast).
    ICCV
    Group-Conditional Conformal Prediction via Quantile Regression Calibration for Crop and Weed Classification
    Paul Melki, Lionel Bombrun, Boubacar Diallo, Jérôme Dias, Jean-Pierre Da Costa
    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023. [Paper]
    This article presents the conformal prediction framework that provides valid statistical guarantees on the predictive performance of any black box prediction machine, with almost no assumptions, applied to the problem of deep visual classification of weeds and crops in real-world conditions.
    ICCV Poster
    Active learning for efficient annotation in crop-weed semantic segmentation.
    Blok, P. M., van Marrewijk, B. M., Boubacar Diallo, Dandjinou, C., Gonzalez, N. F., Melki, P., ... & Dias, J.
    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023. [Paper]
    In agriculture, datasets tend to contain more redundant images and imbalanced classes. Therefore, in this research the added value of active learning was tested on a Corn-Weed dataset. Three acquisition functions were compared: BALD, PowerBALD and Random. Both BALD and PowerBALD outperformed Random sampling even when 90.9% of the pixels belonged to the background class.

    2022

    CEA: Elsevier
    Active learning with MaskAL reduces annotation effort for training Mask R-CNN on a broccoli dataset with visually similar classes
    Pieter M Blok, Gert Kootstra, Hakim Elchaoui Elghor, Boubacar Diallo, Frits K van Evert, E. van Henten.
    Computers and Electronics in Agriculture, Elsevier (CEA: Elsevier), 2022. [Paper]
    The goal of our work was to reduce the number of annotated images needed to train a CNN while maintaining its performance. We use an active learning method capable of automatically selecting difficult-to-classify images and based on the Mask R-CNN instance segmentation algorithm and named this method MaskAL.
    GRETSI
    Analyse exploratoire de métriques en segmentation d'images: application en proxidétection
    Melki, P., Bombrun, L., Millet, E., Boubacar Diallo, Elghor, H. E., & da Costa, J. P.
    In Groupe d'Etudes du Traitement du Signal et des Images (GRETSI), 2022. [Paper]
    Working in an agricultural context, specifically on the problem of the automatic detection of plants in proximal sensing images, we studied twelve evaluation metrics that we used to evaluate three image segmentation models recently presented in the literature. After a unified presentation of these metrics, we carried out an exploratory analysis of their relationships using a correlation analysis, a clustering of variables, and two factorial analyses (namely principal component analysis and multiple factorial analysis). .

    2020

    Ph.D Report
    Mesure de l'intégrité d'une image : des modèles physiques aux modèles d'apprentissage profond
    Boubacar Diallo, HAL science ouverte, December 2020. [Paper]
    In this thesis, we focus on several of these image forensic challenges including camera model identification and image tampering detection. After reviewing the state of the art in the field, we propose a first data-driven method for identifying camera models. We use deep learning techniques based on convolutional neural networks (CNNs) and develop a learning strategy considering the quality of the input data versus the applied transformation. A family of CNN networks has been designed to learn the characteristics of the camera model directly from a collection of images undergoing the same transformations as those commonly used on the Internet.
    FSI
    Robust forgery detection for compressed images using CNN supervision
    Boubacar Diallo, Thierry Urruty, Pascal Bourdon, Christine Fernandez-Maloigne.
    Forensic Science International: Reports (FSI: Reports), 2020. [Paper]
    This article presents a framework improving robustness for image forgery detection. The most important step of our framework is to take into account the image quality corresponding to the chosen application. Therefore, we relied on a camera identification model based on convolutional neural networks.

    2019

    MMM
    Improving robustness of image tampering detection for compression
    Boubacar Diallo, Thierry Urruty, Pascal Bourdon, Christine Fernandez-Maloigne.
    MultiMedia Modeling: 25th International Conference (MMM), 2019. [Paper]
    Improving the robustness of tampered image detection algorithms is essential for compression manipulation as it is the most common type of post-processing on the Internet.