Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FruitProm: Probabilistic Maturity Estimation and Detection of Fruits and Vegetables

Sidharth Rai, Rahul Harsha Cheppally, Benjamin Vail, Keziban Yalçın Dokumacı, Ajay Sharda

TL;DR

The paper addresses maturation estimation for fruits and vegetables by reframing it as a continuous, probabilistic task rather than a discrete classification problem. It introduces FruitProm, which augments the real-time RT-DETRv2 detector with a Probabilistic Maturity Head that predicts a Beta-distributed maturity per detected fruit, along with associated uncertainty. A maturity-aware training objective combines object detection losses with a maturity likelihood term and a label assignment cost that includes maturity, achieving state-of-the-art mAP on a challenging tomato dataset ($85.6%$). The approach yields uncertainty estimates that support robust, autonomous harvesting decisions and improved detection of small or ambiguous fruits, enabling more reliable agricultural robotics.

Abstract

Maturity estimation of fruits and vegetables is a critical task for agricultural automation, directly impacting yield prediction and robotic harvesting. Current deep learning approaches predominantly treat maturity as a discrete classification problem (e.g., unripe, ripe, overripe). This rigid formulation, however, fundamentally conflicts with the continuous nature of the biological ripening process, leading to information loss and ambiguous class boundaries. In this paper, we challenge this paradigm by reframing maturity estimation as a continuous, probabilistic learning task. We propose a novel architectural modification to the state-of-the-art, real-time object detector, RT-DETRv2, by introducing a dedicated probabilistic head. This head enables the model to predict a continuous distribution over the maturity spectrum for each detected object, simultaneously learning the mean maturity state and its associated uncertainty. This uncertainty measure is crucial for downstream decision-making in robotics, providing a confidence score for tasks like selective harvesting. Our model not only provides a far richer and more biologically plausible representation of plant maturity but also maintains exceptional detection performance, achieving a mean Average Precision (mAP) of 85.6\% on a challenging, large-scale fruit dataset. We demonstrate through extensive experiments that our probabilistic approach offers more granular and accurate maturity assessments than its classification-based counterparts, paving the way for more intelligent, uncertainty-aware automated systems in modern agriculture

FruitProm: Probabilistic Maturity Estimation and Detection of Fruits and Vegetables

TL;DR

The paper addresses maturation estimation for fruits and vegetables by reframing it as a continuous, probabilistic task rather than a discrete classification problem. It introduces FruitProm, which augments the real-time RT-DETRv2 detector with a Probabilistic Maturity Head that predicts a Beta-distributed maturity per detected fruit, along with associated uncertainty. A maturity-aware training objective combines object detection losses with a maturity likelihood term and a label assignment cost that includes maturity, achieving state-of-the-art mAP on a challenging tomato dataset (). The approach yields uncertainty estimates that support robust, autonomous harvesting decisions and improved detection of small or ambiguous fruits, enabling more reliable agricultural robotics.

Abstract

Maturity estimation of fruits and vegetables is a critical task for agricultural automation, directly impacting yield prediction and robotic harvesting. Current deep learning approaches predominantly treat maturity as a discrete classification problem (e.g., unripe, ripe, overripe). This rigid formulation, however, fundamentally conflicts with the continuous nature of the biological ripening process, leading to information loss and ambiguous class boundaries. In this paper, we challenge this paradigm by reframing maturity estimation as a continuous, probabilistic learning task. We propose a novel architectural modification to the state-of-the-art, real-time object detector, RT-DETRv2, by introducing a dedicated probabilistic head. This head enables the model to predict a continuous distribution over the maturity spectrum for each detected object, simultaneously learning the mean maturity state and its associated uncertainty. This uncertainty measure is crucial for downstream decision-making in robotics, providing a confidence score for tasks like selective harvesting. Our model not only provides a far richer and more biologically plausible representation of plant maturity but also maintains exceptional detection performance, achieving a mean Average Precision (mAP) of 85.6\% on a challenging, large-scale fruit dataset. We demonstrate through extensive experiments that our probabilistic approach offers more granular and accurate maturity assessments than its classification-based counterparts, paving the way for more intelligent, uncertainty-aware automated systems in modern agriculture

Paper Structure

This paper contains 22 sections, 6 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Uncertainty Quantification in Computer Vision
  4. Probabilistic Prediction Heads for Continuous Attributes
  5. Foundational Architecture: RT-DETR
  6. Backbone and Multi-Scale Feature Extraction
  7. Efficient Hybrid Encoder
  8. Query Selection and Transformer Decoder
  9. Methodology
  10. Model Architecture Modifications
  11. Beta Distribution for Bounded Traits
  12. Probabilistic Prediction Head
  13. Loss function
  14. Label Assignment with Maturity Cost
  15. Final Loss Composition
  16. ...and 7 more sections

Figures (2)

  • Figure 1: Architecture of the probabilistic maturity head integrated into RT-DETRv2. The head predicts Beta distribution parameters ($\alpha,\beta$) per query and is applied to encoder and decoder layers for deep supervision.
  • Figure 2: Qualitative results of FruitProm on tomato image. The predicted maturity distributions illustrate the model's uncertainty handling across different ripening stages.