temp

app.py

@@ -1,23 +1,46 @@
 import gradio as gr
 
-def greet(name):
-    return "Hello " + name + "!!"
+import torch
 
+from how.networks import how_net
 
-# Model to use
-net_path = 'fire.pth'
+import fire_network
 
-# CPU / GPU
-device = 'cpu'
 
-# Images will be downscaled to this size prior processing with the network
-image_size = 1024
+# Possible Scales for multiscale inference
+    scales = [2.0, 1.414, 1.0, 0.707, 0.5, 0.353, 0.25] 
+    infer_opts = {"scales": scales, "features_num": 1000}
+
+
+# Load net
+state = torch.load('fire.pth', map_location='cpu')
+state['net_params']['pretrained'] = None # no need for imagenet pretrained model
+net = fire_network.init_network(**state['net_params']).to(device)
+net.load_state_dict(state['state_dict'])
+
+transforms_ = transforms.Compose([
+        transforms.Resize(1024),
+        transforms.ToTensor(), 
+        transforms.Normalize(**dict(zip(["mean", "std"], net.runtime['mean_std'])))
+        ])
+
 
 # Wrapper
 def generate_matching_superfeatures(im1, im2, scale=6):
     
-    # Possible Scales for multiscale inference
-    scales = [2.0, 1.414, 1.0, 0.707, 0.5, 0.353, 0.25] 
+    # extract features
+    with torch.no_grad():
+            output1 = net.get_superfeatures(im1.to(device), scales=scales)
+            feats1 = output1[0]
+            attns1 = output1[1]
+            strenghts1 = output1[2]
+
+            output2 = net.get_superfeatures(im2.to(device), scales=scales)
+            feats2 = output2[0]
+            attns2 = output2[1]
+            strenghts2 = output2[2]
+
+
 
 
 # GRADIO APP
@@ -31,12 +54,13 @@ iface = gr.Interface(
     inputs=[
         gr.inputs.Image(shape=(240, 240), type="pil"),
         gr.inputs.Image(shape=(240, 240), type="pil"),
-        gr.inputs.Slider(minimum=1, maximum=7, step=1, default=2, label="Scale")],
+        gr.inputs.Slider(minimum=1, maximum=7, step=1, default=2, label="Scale"),
+        gr.inputs.Slider(minimum=1, maximum=255, step=25, default=50, label="Binarizatio Threshold")],
     outputs="plot",
     enable_queue=True,
     title=title,
     description=description,
     article=article,
-    examples=[["chateau_1.png", "chateau_2.png", 6]],
+    examples=[["chateau_1.png", "chateau_2.png", 6, 50]],
 )
 iface.launch()

fire_network.py

@@ -6,8 +6,6 @@ import torch
 from torch import nn
 import torchvision
 
-from cirtorch.networks import imageretrievalnet
-
 from how import layers
 from how.layers import functional as HF
 
@@ -103,7 +101,7 @@ def init_network(architecture, pretrained, skip_layer, dim_reduction, lit, runti
 
     if skip_layer > 0:
         features = features[:-skip_layer]
-    backbone_dim = imageretrievalnet.OUTPUT_DIM[architecture] // (2 ** skip_layer)
+    backbone_dim = 2048 // (2 ** skip_layer)
 
     att_layer = layers.attention.L2Attention()
 

requirements.txt

@@ -3,3 +3,4 @@ pyaml
 matplotlib
 torch==1.10.2
 torchvision==0.11.3
+opencv-python=4.5.5