Two-level classification using single model

Question

At first, i want to binary classify "Fire" event in 5000 images. Secondly, if fire is available in images, then classify further either its urban area(buildings) or rural (As forest). I am using Transfer learning with different models including VGG16 and fine-tune few of its last layers.

I have already tried by training and testing both classification steps separately, but it creates lot of penalty, if i identify that image has fire in rural area but image does not include fire.

I want transfer learning model to binary classify and produce results of both steps as:

img1 fire rural/urban 1 No-fire No-rural/no-urban 2 Fire urban 3 Fire rural

so can i retrain vgg16 in a way to it provides both level of classification i-e step one (fire/no-fire) and step two (rural/urban)

How a same model can be used for two level classification? Anyways, i found the answer. — Muhammad Hanif, Jul 08 '19 at 11:18

score 0 · Answer 1 · answered Jul 08 '19 at 11:20

0

I have found the answer, the same model will be trained to binary classify availability of fire and its weights will be saved. On next step, same weights can be used for second step, to identify either fire is in rural areas or urban areas. Thanks

answered Jul 08 '19 at 11:20

Muhammad Hanif

26
2

Two-level classification using single model

1 Answers1