added complex token -> patch transformation

Readme.md

@@ -44,10 +44,14 @@ git clone https://github.com/juhongm999/hsnet.git
 
 ### Weights
 
-The MIT license does not apply to these weights.
+The MIT license does not apply to these weights. 
+
+We provide two model weights, for D=64 (4.1MB) and D=16 (1.1MB).
+```
+wget https://owncloud.gwdg.de/index.php/s/ioHbRzFx6th32hn/download -O weights.zip
+unzip -d weights -j weights.zip
+```
 
-- [CLIPSeg-D64](https://github.com/timojl/clipseg/raw/master/weights/rd64-uni.pth) (4.1MB, without CLIP weights)
-- [CLIPSeg-D16](https://github.com/timojl/clipseg/raw/master/weights/rd16-uni.pth) (1.1MB, without CLIP weights)
 
 ### Training and Evaluation
 

models/clipseg.py

@@ -300,7 +300,7 @@ class CLIPDensePredT(CLIPDenseBase):
     def __init__(self, version='ViT-B/32', extract_layers=(3, 6, 9), cond_layer=0, reduce_dim=128, n_heads=4, prompt='fixed', 
                  extra_blocks=0, reduce_cond=None, fix_shift=False,
                  learn_trans_conv_only=False,  limit_to_clip_only=False, upsample=False, 
-                 add_calibration=False, rev_activations=False, trans_conv=None, n_tokens=None):
+                 add_calibration=False, rev_activations=False, trans_conv=None, n_tokens=None, complex_trans_conv=False):
 
         super().__init__(version, reduce_cond, reduce_dim, prompt, n_tokens)
         # device = 'cpu'
@@ -337,7 +337,22 @@ def __init__(self, version='ViT-B/32', extract_layers=(3, 6, 9), cond_layer=0, r
             # explicitly define transposed conv kernel size
             trans_conv_ks = (trans_conv, trans_conv)
 
-        self.trans_conv = nn.ConvTranspose2d(reduce_dim, 1, trans_conv_ks, stride=trans_conv_ks)
+        if not complex_trans_conv:
+            self.trans_conv = nn.ConvTranspose2d(reduce_dim, 1, self.trans_conv_ks, stride=self.trans_conv_ks)
+        else:
+            assert self.trans_conv_ks[0] == self.trans_conv_ks[1]
+
+            tp_kernels = (self.trans_conv_ks[0] // 4, self.trans_conv_ks[0] // 4)
+
+            self.trans_conv = nn.Sequential(
+                nn.Conv2d(reduce_dim, reduce_dim, kernel_size=3, padding=1),
+                nn.ReLU(),
+                nn.ConvTranspose2d(reduce_dim, reduce_dim // 2, kernel_size=tp_kernels[0], stride=tp_kernels[0]),
+                nn.ReLU(),
+                nn.ConvTranspose2d(reduce_dim // 2, 1, kernel_size=tp_kernels[1], stride=tp_kernels[1]),               
+            )
+
+#        self.trans_conv = nn.ConvTranspose2d(reduce_dim, 1, trans_conv_ks, stride=trans_conv_ks)
 
         assert len(self.extract_layers) == depth