sbrl-github/research-rainfallradar
1
0
Fork 0
mirror of https://github.com/sbrl/research-rainfallradar synced 2024-11-22 01:12:59 +00:00
Code Issues Projects Releases Packages Wiki Activity

deeplabv3+: prepare for ConvNeXt

Browse source
This commit is contained in:
Starbeamrainbowlabs 2023-03-14 21:51:41 +00:00
parent 779b546897
commit e565c36149
Signed by: sbrl
GPG key ID: 1BE5172E637709C2
1 changed files with 13 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
aimodel/src/deeplabv3_plus_test_rainfall.py
View file

@ -135,21 +135,27 @@ if PATH_CHECKPOINT is None:
return output
def DeeplabV3Plus(image_size, num_classes, num_channels=3):
def DeeplabV3Plus(image_size, num_classes, num_channels=3, backbone="resnet"):
model_input = tf.keras.Input(shape=(image_size, image_size, num_channels))
x = tf.keras.layers.UpSampling2D(size=2)(model_input)
resnet50 = tf.keras.applications.ResNet50(
match backbone:
case "resnet":
backbone = tf.keras.applications.ResNet50(
weights="imagenet" if num_channels == 3 else None,
include_top=False, input_tensor=x
)
x = resnet50.get_layer("conv4_block6_2_relu").output
case _:
raise Exception(f"Error: Unknown backbone {backbone}")
x = backbone.get_layer("conv4_block6_2_relu").output
x = DilatedSpatialPyramidPooling(x)
input_a = tf.keras.layers.UpSampling2D(
size=(image_size // 4 // x.shape[1] * 2, image_size // 4 // x.shape[2] * 2), # <--- UPSAMPLE after pyramid
interpolation="bilinear",
)(x)
input_b = resnet50.get_layer("conv2_block3_2_relu").output
input_b = backbone.get_layer("conv2_block3_2_relu").output
input_b = convolution_block(input_b, num_filters=48, kernel_size=1)
x = tf.keras.layers.Concatenate(axis=-1)([input_a, input_b])