search

DevinBayly / thermal_imaging

access the log plotter here https://devinbayly.github.io/thermal_imaging/
https://devinbayly.github.io/thermal_imaging/
0 stars 0 forks source link

robustness estimates #33

Open DevinBayly opened 2 years ago

DevinBayly commented 2 years ago

This is an issue that serves to explore how the tool is performing with regards to videos that have many video artifacts in them. One example is Camera 3 - 192.168.0.113 (AXIS Q1941-E) - 13-20210528-132219.mp4 from the rockfall clips. Here are screen recordings of how it handles these videos even when they contain lots of pixel noise.

https://user-images.githubusercontent.com/11687631/146826014-031aae48-0733-465c-8f02-8ea3ba39a251.mp4

No centroids are drawn over this in this version.

DevinBayly commented 2 years ago

Here's a really extreme example of the kinds of noise that the code would have to deal with. The original video is shown on the left with some modifications and the right side is the result of the background segmentation and centroid targeting.

https://user-images.githubusercontent.com/11687631/146828785-2211d3ec-209b-4c32-8f2b-055a067a6147.mp4

Overall this shows that when a major artifact frame is encountered it appears as many many centroids and this should allow the code to conditionally evaluate whether to believe that these detections are at all meaningful.