This is a follow-up question from my previous post here about explosion detection. I gathered a dataset of explosions. As I'm new to Deep Learning in Keras, I'm trying to see what architecture best suits this problem, given that here we have a cloud of smoke/fire as opposed to an object. Any suggestions?

For instance, I've learned about Faster RCNN or RetinaNet, but that is mostly for object detection. Is it going to be better than say a basic ResNet50? And here real-time prediction requirements are not an issue. So shall I assume a heavier model (e.g. NASNet Large or a Resnet-152 model) is better than a basic ResNet-50 model?

  • $\begingroup$ "I gathered a dataset of explosions", have you found it on the web or have you manually created it? If you found it on the web, maybe you can give an answer to your other question. $\endgroup$ – nbro Nov 13 '19 at 13:01
  • $\begingroup$ No I created a small dataset myself from youtube! Had a hard time. $\endgroup$ – Tina J Nov 13 '19 at 17:54
  • $\begingroup$ In case you do not find a specific architecture suited for your problem, I would suggest you start with the simplest architecture. If it doesn't perform well enough, try with a more complex one, and so on. $\endgroup$ – nbro Nov 14 '19 at 2:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.