Yuma myotis · bat box
A bat dropping out of the box, filmed close in against open sky.
Show the settings (answer key)
Real emergences to practice on. Each dataset is a clipped recording paired with the
.yml settings file that was used on it, so you can work the video yourself and then
compare against a known answer. They span the range from a clean, predictable box emergence to a
building where you have to find the exit before you can count it.
Settings can also be drilled on their own in the module quizzes.
Download every dataset · 75 MB or take them one at a time below
A bat dropping out of the box, filmed close in against open sky.
Bats emerging from a cave with a vertical entrance.
In this video, the key setting to focus on is the bounding box polygon tool. As this cave exit is not a uniform rectangle, better boundaries can be drawn using the polygon tool. Additionally, the user can increase target detection performance by using the Multiple Sizes tool, as this video contains both large and small bats.
Gray bats emerging from a cold cave.
In this video example, the key setting that will make the overall processing of this kind of video perform better is Target Diameter. If you scroll through this video, you will notice both large and small targets. This is where the Multiple Sizes tool comes in handy, to reduce the number of tracks created on one detection and reduce the number of smaller detections being lost.
Bats leaving a house — out of the vent on the side, the front, and a few from the side window.
A “bat-nado” — free-tailed bats swirling up in a column from under the causeway.
In this video, the key settings to focus on are max density, sensitivity, and the exclusion mask. As this is a high-density exit, drawing a large bounding box and masking the majority of the space within it — leaving a small buffer before the edges of the gate — improves detection quality and reduces noise detections. For the high volume of bats, a low sensitivity with max density set high improves detection success. With high-volume videos it is often better to mask out most of the cloud until it reaches the outer margins of the frame, before it hits the bounding box edges: that reduces processing time and improves detection quality.
No datasets match that combination yet. Try “All”.