Caranha 8 cm Dataset

Published: 3 September 2026| Version 1 | DOI: 10.17632/9ry9s4kkj7.1
Contributors:
Sergio Novak,
,
,
,
,
,
,
,
,

Description

The Caranha 8 cm Dataset is a curated repository of annotated images designed for the development and benchmarking of Artificial Intelligence models in precision aquaculture, particularly for fish detection and counting tasks annotated using the Bounding Box method. The dataset contains images of Caranha (Piaractus brachypomus) at the 8 cm growth stage captured in a controlled environment, which enhances visual contrast and facilitates object detection and segmentation processes.The images were obtained under consistent lighting and background conditions, minimizing environmental variability while still preserving common computer vision challenges such as:Variations in orientation and positioning.Subtle scale differences among the 8 cm specimens.Visual similarity between instances (low inter-class variance).All images are annotated with bounding boxes in YOLO format, supporting direct use in state-of-the-art object detection frameworks. This dataset is highly recommended for training and evaluating deep learning architectures, including the YOLO family (v8 to v11) and Faster R-CNN, particularly in scenarios requiring precise counting and detection under controlled densities and specific growth stages.textdataset/ ├── data.yaml ├── train/ │ ├── images/ (617 jpg files) │ └── labels/ (617 txt files | 3,785 bounding boxes) ├── valid/ │ ├── images/ (46 jpg files) │ └── labels/ (46 txt files | 309 bounding boxes) └── test/ ├── images/ (46 jpg files) └── labels/ (46 txt files | 297 bounding boxes)

Files

Steps to reproduce

Download and unzip the dataset. Install Python 3.10 and Ultralytics YOLOv11. Run: pip install ultralytics Use the provided data.yaml file. Train the model: yolo detect train data=data.yaml model=yolov11n.pt epochs=600 imgsz=640 Evaluate performance: yolo detect val model=runs/detect/train/weights/best.pt

Institutions

Categories

Computer Vision, Fish, Aquaculture, Object Detection, YOLOv7

Funders

Licence