A dataset supporting a non-invasive pipeline for automatic fish length estimation using an underwater Stereo-Vision System (SVS). The method combines 2D segmentation masks with 3D scene reconstruction and multi-object tracking, achieving a Mean Absolute Error of 1.30 cm and a Mean Absolute Percentage Error of 4.53%. The data was authored by Caterina Muntaner-Gonzalez and last updated on 2026-05-25.
Use Cases
- Training and evaluating fish detection models based on the described YOLOv11-based segmentation.
- Developing 3D reconstruction algorithms for underwater scenes using synchronized stereo image pairs.
- Improving multi-object tracking for marine species based on the BoT-SORT tracker mentioned.
- Benchmarking non-invasive fish length estimation methods under challenging underwater conditions.
Strengths
- The pipeline achieved a Mean Absolute Error (MAE) of 1.30 cm and a Mean Absolute Percentage Error (MAPE) of 4.53%.
- Methodology includes a multi-object tracking module for track-based aggregation to improve robustness.
- The pipeline is integrated into an autonomous system enabling real-time on-board operation.
Limitations
- The dataset size is 78.0 B, indicating a very limited scope.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- figshare
- Collection Method
- Collected via an autonomous underwater Stereo-Vision System (SVS) providing synchronized image pairs for 3D reconstruction.
- Freshness
- Last updated 2026-05-25 04:33:16; freshness should be verified.
- Geography
- Marine environments; specific locations are not stated.