Update app.py
Browse files
app.py
CHANGED
@@ -191,14 +191,13 @@ train_images = get_sample_images(TRAIN_FOLDER)
|
|
191 |
test_images = get_sample_images(TEST_FOLDER)
|
192 |
|
193 |
description= '''
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
validation accuracy of about 75% and dice score of about 0.6.
|
202 |
'''
|
203 |
|
204 |
# Create the train dataset interface
|
|
|
191 |
test_images = get_sample_images(TEST_FOLDER)
|
192 |
|
193 |
description= '''
|
194 |
+
Satellite imagery, powered by advancements in computer vision and GPU technology, plays a crucial
|
195 |
+
role in urban planning and climate change research. Deep learning, particularly the U-Net model,
|
196 |
+
automates land cover classification and facilitates monitoring of seagrass habitats in the Mediterranean Sea.
|
197 |
+
Our project, using the DeepGlobe Land Cover Classification Challenge 2018 dataset, trained four models
|
198 |
+
(basic U-Net, VGG16 U-Net, Resnet50 U-Net, and Efficient Net U-Net) and
|
199 |
+
achieved a validation accuracy of approximately 75% and a dice score of about 0.6 through an ensemble
|
200 |
+
approach on 803 images with segmentation masks (80/20 split).
|
|
|
201 |
'''
|
202 |
|
203 |
# Create the train dataset interface
|