Deep learning offers a non-destructive route to automated fruit-quality screening

The gap

Evidence profile

Sourced from the future work and inline gaps and stated research gap and abstract of the source papers, classified as general, spanning 6 journals.

Supporting evidence — 8 representative gaps

• Deep learning and IoT-based framework for sesame plant identification and weed detection (2026) · Scientific Reports · doi

  Identifying weeds correctly is important for keeping plants healthy in the future and getting the most out of your farm. This paper presents a novel methodology for weed identification, leveraging the capabilities of deep learning algorithms to facilitate streamlined weed detection. We used a tailored model with an RCNN architecture. It can help make boundary boxes for both weeds and crops, which makes it easier to tell the difference between ARTICLE IN PRESSARTICLE IN PRESS ACCEPTED MANUSCRIPT the two. The analysis shows that it works very well because the highest bounding box accuracy is 0.720, the lowest bounding box accuracy is always 0.221, and the validation class accuracy can be as high as 0.950. The suggested procedures make it easy to accurately identify weeds in crop fields. The dataset (around 1300 photos) was carefully labeled and shows typical field situations, making it a good proof-of-concept. The results show how region-based CNNs and advanced object detection techniques may work together for precision agriculture. Future work will explore one-stage detectors such as YOLO for deployment-critical scenarios.The possibility of employing a Region Proposal Network (RPN) in lieu of selective search presents a prospective avenue for future research, highlighting its capacity to enhance runtime efficiency. Modern end-to-end models outperform CNN+SVM in speed, but the proposed approach offers analytical flexibility. By adding IoT sensor data, the system may be made to give irrigation advice that takes into account the moisture level of the soil. Also, using Internet of Things sensors in the last few years will make this method even better at assessing soil moisture levels and pest populations, which will improve crop management even more. Conflict of interests: There is no conflict of interests among students: Authors contributions: P.N. and V.M. conceived the study and designed the methodology. K.M.S. implemented the software, performed the formal analysis, and prepared the visualizations. S.A.K.S.A. and S.M. carried out the investigation and curated the data. C.K.C. and C.S. provided resources, contributed to validation, and supervised the work. P.N. and K.M.S. drafted the manuscript, and V.M., S.A.K.S.A., S.M., C.K.C., and C.S. reviewed and edited it. P.N. coordinated the project. All authors approved the final manuscript. Data availability statement: ARTICLE IN PRESSARTICLE IN PRESS ACCEPTED MANUSCRIPT All data used in the manuscript is within the manuscript itself. Acknowledgement: NIL Funding: No Funding is received.

  generalfuture work
  Keywords: manuscript article press weeds future make accuracy presents methodology weed detection used accepted shows bounding
• AI-Powered Plant Detector: A Cloud-Based Plant Disease Detectionand Classification System (2026) · International Journal of Engineering Technology and Management Sciences · doi

  6.1 Conclusion The AI-Powered Plant Detector is a comprehensive and intelligent cloud-based web application that successfully integrates Artificial Intelligence, Deep Learning, and Cloud Computing to revolutionize the way individuals monitor and manage plant health. Throughout this project, a fully functional system was designed, developed, tested, and deployed that automates plant disease detection, intelligently classifies diseases using a CNN-based Deep Learning Model with an accuracy of 94.7%, and presents meaningful agricultural insights through an interactive and visually rich dashboard. The system achieved a disease alert accuracy of 96%, a system uptime of 99.7%, and an average user satisfaction score of 4.6 out of 5, with users reporting a 40% improvement in agricultural awareness after just two weeks of usage.

  generalfuture work
  Keywords: plant system cloud based deep learning disease accuracy agricultural conclusion powered detector comprehensive intelligent application
• Intelligent Mushroom Classification with Machine Learning and Deep Learning: A Comprehensive Survey and Future Directions (2026) · Journal of Macrofungi · doi

  To overcome these challenges, future research may consider the following directions: ▪ Real-time mobile recognition apps: Development of lightweight CNNs, MobileNets, or YOLO-based detec- tors optimized for smartphones and IoT platforms will enable farmers, food inspectors, and consumers to identify mushrooms instantly in the field. ▪ Hybrid models (ML+expert systems): Combining data-driven ML/DL methods with expert knowledge (rules, ontologies, or symbolic AI) could yield more interpretable & reliable systems for high-risk edibility prediction. ▪ Expansion of fungal databases with AI integration: Creating and sharing large, standardized fungal image and attribute repositories, enriched with metadata (geolocation, environmental conditions), will facilitate cross-dataset training and benchmarking. AI can further assist in automated annotation and validation. ▪ Multimodal fusion and domain adaptation: Integrating images, sensor data (humidity, soil conditions), and morphological attributes, while employing domain adaptation methods, can help models generalize across unseen environments and species. ▪ Explainability and transparency: Building explainable AI tools (e.g., saliency maps, interpretable ML) will improve trust in real-world applications, particularly in food safety and medical contexts. 7. CONCLUSION This review has offered a broad integration of ML and DL methods used on mushroom classification and edibility. The ease of learning has enabled classical ML algorithms, including DT, RF, and SVM, to narrow their accuracy gap against the tabular datasets to perfect values, while DL architectures such as CNNs, transfer learning models, and YOLO-based detectors have achieved impressive results on image classification and real-time detection. Hybrid methods, where the deep features are fused with classical classifiers, lead to even more significant improvements when data is limited. However, problems still remain, such as inadequate annotated data sets, unbalanced classes and a lack of generalization to practical scenarios. Solving these challenges will necessitate a collective undertaking in dataset curation, multimodal data integration, and the development of light-weight, explainable, and real-time systems. ML and DL have an enormous potential in improving systems for recognizing mushrooms. With further development in multimodal learning, dataset growth, and AI-enabled mobile deployment, the next generation of intelligent tools may be transformative for agriculture, ecology, and food safety.

  generalfuture work
  Keywords: real systems time development food models integration dataset multimodal learning challenges mobile cnns yolo based
• AI-based visual analysis of photovoltaic panels for fault detection and maintenance support (2026) · Scientific Reports · doi

  In light of our experimental observations, the primary focus of future work will involve expanding the current dataset to include underrepresented defect categories that were identified as challenging during the training phase. Since our results indicated that certain visual similarities between texture-based contamination and physical damage lead to occasional misclassifications, we intend to refine the feature extraction process to capture more granular textural details. Furthermore, the variability observed across the cross-validation folds suggests a need for more robust data augmentation techniques that can better represent variations in lighting conditions and camera angles. Beyond improving standalone classification accuracy, future research should prioritize the integration of these deep learning models into automated aerial inspection frameworks. The transition from offline diagnostic tools to real-time monitoring systems requires the development of lightweight architectures that can maintain high precision on resource-constrained edge devices, such as drones. Future efforts will therefore focus on model compression and quantization techniques to ensure operational feasibility in large-scale PV plants38. Additionally, incorporating multi-spectral or temporal data could provide a more comprehensive diagnostic foundation to effectively distinguish between transient surface anomalies and permanent structural defects.

  generalfuture work
  Keywords: future focus techniques diagnostic light experimental observations primary involve expanding current dataset include underrepresented defect
• Image Processing-Based Detection of Pomegranate Leaf Diseases Using K-Means Clustering and SVM (2026) · International Journal of Creative and Open Research in Engineering and Management · doi

  Future work will focus on expanding the dataset to include a wider variety of leaf disease stages and environmental conditions, integrating deep learning architectures such as Convolutional Neural Networks (CNNs) to improve detection robustness and accuracy, and deploying the system as a lightweight mobile or IoT-based application to enable real-time, field-level leaf disease diagnosis. The findings of this study confirm that image processing and machine learning techniques offer a viable and effective alternative to conventional manual leaf inspection methods, which are time- field varying conditions, Despite its promising performance, the system has certain limitations that warrant further investigation.

  generalinline gaps
  Keywords: leaf disease conditions learning system time field future focus expanding dataset include wider variety stages
• Explainable Hybrid CNN-Transformer Frameworkfor Papaya Leaf Disease Classification withLayer-Wise Grad-CAM Analysis (2026) · AIUB Journal of Science and Engineering (AJSE) · doi

  The lack of transparency and interpretability in deep learning models for plant disease detection limits their adoption by agricultural practitioners. The need for a systematic approach to capturing progressive feature evolution from edge detection through texture analysis to disease-region localization. The limited evaluation of deep learning models for plant disease detection in real-world field settings.

  generalstated research gapevidence 5/5
  Keywords: lack transparency interpretability deep learning models plant disease
• Performance evaluation of AkidaNet converted to spiking domain for the classification of weeds in cotton fields (2026) · Scientific Reports · doi

  There is a gap in the literature for efficient and accurate deep learning models that can be deployed on edge devices and neuromorphic hardware. Prior work has focused on using deep learning techniques such as YOLOv8n and ResNet101, but these models often require significant computational resources. There is a need for lightweight and efficient deep learning models that can be used for real-time image classification tasks in precision agriculture.

  generalstated research gapevidence 5/5
  Keywords: there gap literature efficient accurate deep learning models
• A Multi-Class Deep Learning Architecture for Quality Classification and Formalin Contamination Detection in Horticultural Fruits (2026) · Jurnal Inotera · doi

  Deep learning offers a non-destructive route to automated fruit-quality screening; however, the literature has largely pursued peak accuracy using progressively heavier network backbones, leaving the computational cost of that accuracy and therefore the feasibility of on-site, embedded deployment insufficiently examined.

  generalabstractevidence 5/5
  Keywords: accuracy deep learning offers destructive route automated fruit quality screening literature largely pursued peak using

Questions about this gap