Employing CNN mobileNetV2 and ensemble models in classifying drones forest fire detection images
Bibliographic record
Abstract
In recent years, the adoption of advanced machine learning techniques has revolutionized approaches to solving complex problems, such as identifying occurrences of forest fires. Among these techniques, the use of Convolutional Neural Networks (CNNs) combined with ensemble methods is particularly promising. To investigate the feasibility of detecting fires using video streams from Unmanned Aerial Vehicles (UAVs), the lightweight CNN architecture MobileNetV2 was utilized for real-time detection. Several experiments were conducted on the DeepFire dataset, which comprises an equal number of images with and without fire, to evaluate MobileNetV2's performance. Notably, the architecture's linear bottlenecks and the efficient use of inverted residuals ensure high accuracy without compromising on feature extraction capabilities. For a comprehensive assessment, MobileNetV2 was benchmarked against other models, including DenseNet121, EfficientNetV2S, and VGG16. Accuracy was enhanced by averaging predictions through methods such as voting or summing results. As documented in the literature, MobileNetV2 consistently outperforms other architectures in computational efficiency and provides an excellent balance between efficiency and the quality of learned features over multiple epochs. This study underscores the suitability of MobileNetV2 for real-time applications on drones, particularly for the detection of forest fires in resource-constrained environments. The results show that MobileNetV2 achieves the highest accuracy (0.994), sensitivity (0.994), and specificity (0.998) among the tested models, with low standard deviations across all metrics. In contrast, EfficientNetV2S exhibited the lowest accuracy and sensitivity, both at 0.779, with a specificity of 0.829. The ensemble (Sum) method achieved an average accuracy of 0.989, sensitivity of 0.989, and specificity of approximately 0.988. Therefore, MobileNetV2 not only delivers the highest accuracy and stability but also demonstrates that the choice of ensemble method significantly affects the results.
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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".