vapaau/3D-ZeF

This is the dataset page for 3D-ZeF, the first RGB 3D multiple object tracking dataset of its kind. Zebrafish is a widely used model organism for studying neurological disorders, social anxiety, and more. Behavioral analysis can be a critical part of such research and it has traditionally been conducted manually, which is an expensive, time consuming, and subjective task.

Dataset Setup

We have used an off-the-shelf setup for capturing the dataset, which consists of two GoPro cameras, a glass aquarium, and two IKEA light panels. We have used this relatively cheap setup on purpose in order to hopefully make automated 3D behavioral analysis of zebrafish more accessible to smaller labs, students, and the likes.

Dataset Files

The dataset consists of a 8 sequences, with each sequence consists of two recordings, one from each of the two GoPros along with ground truth labels and calibration files.

All the fish have been manually bounding and point annotated in all frames. The point annotations are placed on the head of the fish in order to have a stable point for the stereo 3D reconstruction phase. In cases where a fish head is occluded we have estimated the head point annotations as best as possible and only the visible part of the fish has been bounding box annotated. In cases where the fish is totally occluded the bounding box will only be an estimated point.

The sequences are split into a training, validation, and test split. Each split consists of multiple sequences and each sequence consists of two recordings. It should be noted that unique fish are used in each split. Furthermore, the fish in the training and validation split are from one cohort of fish and the zebrafish in the test split are from a completely different and younger cohort. This is to avoid data leakage and overfitting.

Train Split: ZebraFish-01, ZebraFish-02

Validation Split: ZebraFish-03, ZebraFish-04

Test Split: ZebraFish-05, ZebraFish-06, ZebraFish-07, ZebraFish-08

Evaluation

In order to evalaute a tracker performance we recommend using the official MOTChallenge Evaluation Kit: https://github.com/dendorferpatrick/MOTChallengeEvalKit/tree/master/ZF3D

Citation:

@InProceedings{Pedersen_2020_CVPR,
author = {Pedersen, Malte and Haurum, Joakim Bruslund and Bengtson, Stefan Hein and Moeslund, Thomas B.},
title = {3D-ZeF: A 3D Zebrafish Tracking Benchmark Dataset},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}