Camera traps are a non-invasive and time-efficient tool for collecting data on the activity and identity of wildlife at specific locations. A single researcher can easily sample continuously at dozens of locations simultaneously. Camera-trapping surveys, however, come with a new challenge, namely the processing of the enormous volumes of imagery that they may yield. Date, time and technical data can be pulled from the EXIF of the images, but determining what the images contain often remains a burdening task.
One obvious solution is to use deep learning to detect and identify animals in these images.
You will develop one or more algorithms to detect and identify animals within image sequences recorded by the camera traps at National Park Hoge Veluwe (NPHV). These cameras take series of ten images upon every trigger of an infrared sensor, and can immediately be triggered again, yielding sequences of 10, 20, 30 or more images. Thus, algorithms yield multiple classifications for each sequence, which must be aggregated into a single observation. Difficulties are that the images contain one to many individuals, sometimes more than one species, are very different between day (color) and night (IR-B/W), and are sometimes empty. We ask you to try a variety of approaches.
You will be given a library of images for the eight most common species of NPHV. The images have already been separated into day and night frames, and into empty frames and frames with one or more animals.
e.g.:
Marco Willi, Ross T. Pitman, Anabelle W. Cardoso, Christina Locke, Alexandra Swanson, Amy Boyer, Marten Veldthuis & Lucy Fortson (2019). Identifying animal species in camera trap images using deep learning and citizen science. Methods in Ecology and Evolution 10: 80-91.
Danielle L. Norman, Philipp H. Bischoff, Oliver R. Wearn, Robert M. Ewers, J. Marcus Rowcliffe, Benjamin Evans, Sarab Sethi, Philip M. Chapman & Robin Freeman (2022). Can CNN-based species classification generalise across variation in habitat within a camera trap survey? Methods in Ecology and Evolution 14: 242-251.
Veronika Mitterwallner, Anne Peters, Hendrik Edelhoff, Gregor Mathes, Hien Nguyen, Wibke Peters, Marco Heurich & Manuel J. Steinbauer (2024). Automated visitor and wildlife monitoring with camera traps and machine learning. Remote Sensing in Ecology and Conservation 10: 236-247.