search

tobybreckon / fire-detection-cnn

real-time fire detection in video imagery using a convolutional neural network (deep learning) - from our ICIP 2018 paper (Dunnings / Breckon) + ICMLA 2019 paper (Samarth / Bhowmik / Breckon)
MIT License
542 stars 172 forks source link

Fire Annotation #42

Closed multierection closed 4 years ago

multierection commented 4 years ago

If I want to annotate the fire position and get the coordinates from the video using FireNet or InceptionV1-OnFire, what should I do? Thanks !

tobybreckon commented 4 years ago

Use the superpixel versions of the network (which are slower), create a binary image from the output of the network where the regions where the superpixels are Fire are > 0, and everywhere else is 0 (use logical operations, will be most efficient) - then use boundingRect() from OpenCV on this image to get the rectangle coordinates of the fire.

Alternatively just take the min/max x and y of the set of superpixels returned as fire.