Release date
28th July 2026
Changes
Added 2,849 Reactive Learning images to the dataset. These are false positives and false negatives identified by users and internal algorithms. Adding them follows a human-in-the-loop approach to address recurrent field issues and improve model robustness.
Model evaluation
Evaluation metrics
Model 4.32.0 is evaluated against the baseline model 4.31.0 on the same test set. The table below reports AP@0.5 and F0.45 for the person, safety cone and vehicle classes.
AI model version | AP@0.5 (person) | AP@0.5 (safety cone) | AP@0.5 (vehicle) | F0.45 (person) | F0.45 (safety cone) | F0.45 (vehicle) |
4.32.0 | 0.9280 | 0.8527 | 0.8086 | 0.9484 | 0.9152 | 0.8871 |
4.31.0 | 0.9276 | 0.8600 | 0.8090 | 0.9480 | 0.9144 | 0.8871 |
QA Metrics
Both the current and baseline models are evaluated on 817 videos (roughly 261,000 labeled frames) to simulate real-world performance for the person class.
Classification accuracy measures the proportion of frames that were correctly identified as containing people or not containing people.
| 4.31.0 | 4.32.0 | %change |
False positive occurrences | 45 | 34 | -24.44% |
Classification accuracy | 0.9677 | 0.9671 | -0.06% |