Improved YOLO-based real-time brinjal detection algorithm for vision modules in harvesting robots
Bibliographic record
Abstract
Abstract A novel, lightweight, and accurate brinjal detection algorithm, YOLOv11s-Brinjal, was developed for vision modules in selective harvesting robots operating under complex horticultural environments. The algorithm addressed critical detection challenges, including variable lighting, spotlight effects, object overlap, occlusion, and cluttered backgrounds in unstructured farm settings. Multiple configurations from YOLOv8 to YOLOv12 were initially evaluated using a custom dataset, manually annotated and augmented through the Roboflow framework. The best-performing base model, YOLOv11s, was further optimized via systematic channel dimension pruning applied to the convolutional layers of its backbone architecture, significantly reducing both parameter count and computational load. To mitigate performance degradation and ensure task-specific alignment, weight adjustment techniques were implemented during fine-tuning. The YOLOv11s-Brinjal model was evaluated using the same test datasets, demonstrating robust performance with precision, recall, F1 score, and mean average precision values of 94%, 96.6%, 95.3%, and 98.1%, respectively. To assess generalization and detect potential overfitting, a 5-fold cross-validation was conducted. Compared to the original model, the proposed pruning and weight adjustment techniques improved recall by 1.3% , while reducing parameters and computational load by over 57%. With a compact model size of 8.2 MB and an inference time of 10.1 ms, YOLOv11s-Brinjal is well-suited for integration on edge devices as the vision component in real-time selective brinjal harvesting applications.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".