Far infrared images collected from a vehicle driven in outdoor urban scenarios. The dataset was manually annotated with pedestrian bounding boxes and is divided into a classification subset with rescaled images and a detection subset with original images. It was authored by Daniel Olmeda and last updated on May 5, 2024.
Use Cases
- Training pedestrian detection models based on far infrared imagery from urban driving scenarios.
- Developing image classification algorithms based on a dataset of positives and sampled negatives.
- Evaluating object detection performance on manually annotated bounding boxes.
- Researching sensor fusion by comparing FIR data with other modalities like visible light.
- Benchmarking model robustness to conditions where FIR sensors are advantageous.
Strengths
- Images have specific technical parameters: 164x129 pixel resolution and a 14-bit grey-level scale.
- Manual annotation of pedestrians provides a reliable ground truth.
- Dataset is structured for two distinct tasks: classification and detection.
- Data collection method, using an exterior-mounted Indigo Omega imager, aimed to avoid windshield filtering.
Limitations
- Row count and dataset scale are unknown, which may limit suitability assessment.
- Column-level documentation is absent; field semantics must be inferred after download.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- e-cienciaDatos Harvested Dataverse
- Collection Method
- Images acquired with an Indigo Omega imager mounted on a vehicle in urban scenarios and manually annotated.
- Time Range
- null
- Freshness
- Last updated 2024-05-05 07:18:28; freshness should be verified.
- Geography
- Outdoor urban scenarios (specific location not stated).