Algorithm 1.
Proposed CycleGAN-Inception algorithm.
| 1: | x ← input COVID-19 images |
| 2: | y ← input non COVID-19 images |
| 3: | procedure pre-processing (x,y) |
| 4: | x,y ← resize (x,y) |
| 5: | x,y ← normalize (x,y) |
| 6: | end procedure pre-processing |
| 7: | procedure CycleGAN (x,y) |
| 8: | x,y ← G, F |
| 9: | L(G,F,DX,DY)=LGAN(G,DY,X,Y)+LGAN(F,DXY,X)+λLcyc(G,F) |
| 10: | end procedure CycleGAN |
| 11: | procedure Finetune_Inception (n=10, j=50, FC1 = 92, FC2 = 1028, activation = ReLU) |
| 12: | remove FC layers Inception V3 |
| 13: | add_layer (FC1, activation) |
| 14: | add_layer (FC2, activation) |
| 15: | build model CycleGAN_Inception |
| 16: | for k ← 0, n do |
| 17: | for epoch ← 0, j do |
| 18: | train CycleGAN_Inception (x, y, Dx, Dy) |
| 19: | evaluate CycleGAN_Inception (x, y, Dx, Dy) |
| 20: | end for |
| 21: | end for |
| 22: | end procedure Finetune_Inception |