Spaces:
Runtime error
Runtime error
Update glaucoma.py
Browse files- glaucoma.py +45 -7
glaucoma.py
CHANGED
|
@@ -1,28 +1,40 @@
|
|
| 1 |
import cv2
|
| 2 |
import torch
|
| 3 |
|
| 4 |
-
|
| 5 |
|
| 6 |
-
from
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
|
| 8 |
|
| 9 |
class GlaucomaModel(object):
|
| 10 |
def __init__(self,
|
| 11 |
cls_model_path="pamixsun/swinv2_tiny_for_glaucoma_classification",
|
|
|
|
| 12 |
device=torch.device('cpu')):
|
| 13 |
# where to load the model, gpu or cpu ?
|
| 14 |
self.device = device
|
| 15 |
-
#
|
| 16 |
self.cls_extractor = AutoImageProcessor.from_pretrained(cls_model_path)
|
| 17 |
self.cls_model = Swinv2ForImageClassification.from_pretrained(cls_model_path).to(device).eval()
|
|
|
|
|
|
|
|
|
|
| 18 |
# class activation map
|
| 19 |
self.cam = ClassActivationMap(self.cls_model, self.cls_extractor)
|
| 20 |
|
| 21 |
# classification id to label
|
| 22 |
-
self.
|
|
|
|
|
|
|
| 23 |
|
| 24 |
-
# number of classes
|
| 25 |
-
self.num_diseases = len(self.
|
|
|
|
|
|
|
| 26 |
|
| 27 |
def glaucoma_pred(self, image):
|
| 28 |
"""
|
|
@@ -36,6 +48,29 @@ class GlaucomaModel(object):
|
|
| 36 |
disease_idx = outputs.cpu()[0, :].detach().numpy().argmax()
|
| 37 |
|
| 38 |
return disease_idx
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
def process(self, image):
|
| 41 |
"""
|
|
@@ -46,6 +81,9 @@ class GlaucomaModel(object):
|
|
| 46 |
disease_idx = self.glaucoma_pred(image)
|
| 47 |
cam = self.cam.get_cam(image, disease_idx)
|
| 48 |
cam = cv2.resize(cam, image_shape[::-1])
|
|
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
-
return disease_idx, cam
|
| 51 |
|
|
|
|
| 1 |
import cv2
|
| 2 |
import torch
|
| 3 |
|
| 4 |
+
import numpy as np
|
| 5 |
|
| 6 |
+
from torch import nn
|
| 7 |
+
from transformers import AutoImageProcessor, Swinv2ForImageClassification, SegformerForSemanticSegmentation
|
| 8 |
+
|
| 9 |
+
from lib.cam import ClassActivationMap
|
| 10 |
+
from lib.utils import add_mask, simple_vcdr
|
| 11 |
|
| 12 |
|
| 13 |
class GlaucomaModel(object):
|
| 14 |
def __init__(self,
|
| 15 |
cls_model_path="pamixsun/swinv2_tiny_for_glaucoma_classification",
|
| 16 |
+
seg_model_path='pamixsun/segformer_for_optic_disc_cup_segmentation',
|
| 17 |
device=torch.device('cpu')):
|
| 18 |
# where to load the model, gpu or cpu ?
|
| 19 |
self.device = device
|
| 20 |
+
# classification model for glaucoma
|
| 21 |
self.cls_extractor = AutoImageProcessor.from_pretrained(cls_model_path)
|
| 22 |
self.cls_model = Swinv2ForImageClassification.from_pretrained(cls_model_path).to(device).eval()
|
| 23 |
+
# segmentation model for optic disc and cup
|
| 24 |
+
self.seg_extractor = AutoImageProcessor.from_pretrained(seg_model_path)
|
| 25 |
+
self.seg_model = SegformerForSemanticSegmentation.from_pretrained(seg_model_path).to(device).eval()
|
| 26 |
# class activation map
|
| 27 |
self.cam = ClassActivationMap(self.cls_model, self.cls_extractor)
|
| 28 |
|
| 29 |
# classification id to label
|
| 30 |
+
self.cls_id2label = self.cls_model.config.id2label
|
| 31 |
+
# segmentation id to label
|
| 32 |
+
self.seg_id2label = self.seg_model.config.id2label
|
| 33 |
|
| 34 |
+
# number of classes for classification
|
| 35 |
+
self.num_diseases = len(self.cls_id2label)
|
| 36 |
+
# number of classes for segmentation
|
| 37 |
+
self.seg_classes = len(self.seg_id2label)
|
| 38 |
|
| 39 |
def glaucoma_pred(self, image):
|
| 40 |
"""
|
|
|
|
| 48 |
disease_idx = outputs.cpu()[0, :].detach().numpy().argmax()
|
| 49 |
|
| 50 |
return disease_idx
|
| 51 |
+
|
| 52 |
+
def optic_disc_cup_pred(self, image):
|
| 53 |
+
"""
|
| 54 |
+
Args:
|
| 55 |
+
image: image array in RGB order.
|
| 56 |
+
"""
|
| 57 |
+
inputs = self.seg_extractor(images=image.copy(), return_tensors="pt")
|
| 58 |
+
|
| 59 |
+
with torch.no_grad():
|
| 60 |
+
inputs.to(self.device)
|
| 61 |
+
outputs = self.seg_model(**inputs)
|
| 62 |
+
logits = outputs.logits.cpu()
|
| 63 |
+
|
| 64 |
+
upsampled_logits = nn.functional.interpolate(
|
| 65 |
+
logits,
|
| 66 |
+
size=image.shape[:2],
|
| 67 |
+
mode="bilinear",
|
| 68 |
+
align_corners=False,
|
| 69 |
+
)
|
| 70 |
+
|
| 71 |
+
pred_disc_cup = upsampled_logits.argmax(dim=1)[0]
|
| 72 |
+
|
| 73 |
+
return pred_disc_cup.numpy().astype(np.uint8)
|
| 74 |
|
| 75 |
def process(self, image):
|
| 76 |
"""
|
|
|
|
| 81 |
disease_idx = self.glaucoma_pred(image)
|
| 82 |
cam = self.cam.get_cam(image, disease_idx)
|
| 83 |
cam = cv2.resize(cam, image_shape[::-1])
|
| 84 |
+
disc_cup = self.optic_disc_cup_pred(image)
|
| 85 |
+
vcdr = simple_vcdr(disc_cup)
|
| 86 |
+
_, disc_cup_image = add_mask(image, disc_cup, [0, 1, 2], [[0, 0, 0], [0, 255, 0], [255, 0, 0]], 0.2)
|
| 87 |
|
| 88 |
+
return disease_idx, disc_cup_image, cam, vcdr
|
| 89 |
|