(img, size1, size2, pix2pixsize, model, net_type, pix2pixmodel)
| 1026 | |
| 1027 | # Generate a double-input depth estimation |
| 1028 | def doubleestimate(img, size1, size2, pix2pixsize, model, net_type, pix2pixmodel): |
| 1029 | # Generate the low resolution estimation |
| 1030 | estimate1 = singleestimate(img, size1, model, net_type) |
| 1031 | # Resize to the inference size of merge network. |
| 1032 | estimate1 = cv2.resize(estimate1, (pix2pixsize, pix2pixsize), interpolation=cv2.INTER_CUBIC) |
| 1033 | |
| 1034 | # Generate the high resolution estimation |
| 1035 | estimate2 = singleestimate(img, size2, model, net_type) |
| 1036 | # Resize to the inference size of merge network. |
| 1037 | estimate2 = cv2.resize(estimate2, (pix2pixsize, pix2pixsize), interpolation=cv2.INTER_CUBIC) |
| 1038 | |
| 1039 | # Inference on the merge model |
| 1040 | pix2pixmodel.set_input(estimate1, estimate2) |
| 1041 | pix2pixmodel.test() |
| 1042 | visuals = pix2pixmodel.get_current_visuals() |
| 1043 | prediction_mapped = visuals['fake_B'] |
| 1044 | prediction_mapped = (prediction_mapped + 1) / 2 |
| 1045 | prediction_mapped = (prediction_mapped - torch.min(prediction_mapped)) / ( |
| 1046 | torch.max(prediction_mapped) - torch.min(prediction_mapped)) |
| 1047 | prediction_mapped = prediction_mapped.squeeze().cpu().numpy() |
| 1048 | |
| 1049 | return prediction_mapped |
| 1050 | |
| 1051 | |
| 1052 | # Generate a single-input depth estimation |
no test coverage detected