Animals are hiding in the background? This toolchain tries to preprocess those images to make detections work more reliable.
The main application. The idea is to write a complete flow which focusses on detection of animals in the background.
Most computer vision tasks focus on nice, high quality images. The images are mostly
- taken in good weather conditions.
- having good lightning conditions.
- containing sharp and good visible objects of interest.
- ...
For camera-trap images, this highly differs: Shaky images in bad weather taken using dirty lenses are increasing the effort for good image recognitions.
ShadowWolf tries to deal with those challenges and to detect as many objects of interests as possible.
Several tools for analysis etc.
A fundamental documentation of this work.
As the structure has been changed in August 2023, you have to update the database when re-running the old results with the script. The commands are:
UPDATE detected_image SET source_datagetter = 'get_images' WHERE source_datagetter = 'getImages'
UPDATE simple_label_split SET source_getter = 'get_cut_images' WHERE source_getter = 'getCutImages'
UPDATE simple_label_split SET source_getter = 'get_images' WHERE source_getter = 'getImages'
The everything till here just requires minor changes (c.f. previous point). Afterwards, several major changes were done: The results cannot be reused, the scripts have to be re-run.
- Jens Dede, Sustainable Communication Networks, University of Bremen, 2023