NID Dataset – Noctural Insect Detection Dataset

This webpage contains datasets and supplementary information for the following paper:

Prototype of the moth scanner installed in Jena.

Automated Visual Monitoring of Nocturnal Insects with Light-based Camera Traps.
D. Korsch, P. Bodesheim, G. Brehm, and J. Denzler.

CVPR Workshop on Fine-grained Visual Classification (CVPR-WS). 2022.

Description

Example of the bounding box annotations

The images are captured using a moth scanner prototype developed in the context of the AMMOD project. In the time between June and October 2021, we captured over 27.000 images during 95 nights. For 10 nights (equidistantly selected), we annotated 818 images with bounding boxes describing the position of the captured insects. This resulted in 9.095 annotated insects.

Validation splits

For the evaluation of the detection methods, we split the dataset day-wise. First five days of the annotated images were used for training and the last five for validation. This splitting strategy results in 522 training and 296 validation images with 7.532 and 1.563 annotated insects, respectively.

Ownership

The dataset was acquired in the context of the AMMOD project with a moth scanner prototype in Jena by Gunnar Brehm. The annotations were created by Gunnar Brehm and Dimitri Korsch.

Downloads

There are two versions of the dataset:

To avoid automatic downloads and unreasonable load on our server, you need to enter a password (“nid_download“)