example

examples/hello-world.py

@@ -32,6 +32,7 @@ model = scn.Sequential().add(
 # output will be 10x10
 inputSpatialSize = model.input_spatial_size(torch.LongTensor([10, 10]))
 input_layer = scn.InputLayer(2, inputSpatialSize)
+bl_input_layer = scn.BLInputLayer(2, inputSpatialSize)
 
 msgs = [[" X   X  XXX  X    X    XX     X       X   XX   XXX   X    XXX   ",
          " X   X  X    X    X   X  X    X       X  X  X  X  X  X    X  X  ",
@@ -46,7 +47,7 @@ msgs = [[" X   X  XXX  X    X    XX     X       X   XX   XXX   X    XXX   ",
          " X     X          XXXX   x   x     x   x  xxxx     x ",]]
 
 
-# Create Nx3 and Nx1 vectors to encode the messages above:
+# Create Nx3 and Nx1 vectors to encode the messages above using InputLayer:
 locations = []
 features = []
 for batchIdx, msg in enumerate(msgs):
@@ -65,3 +66,30 @@ output = model(input)
 # Output is 2x32x10x10: our minibatch has 2 samples, the network has 32 output
 # feature planes, and 10x10 is the spatial size of the output.
 print('Output SparseConvNetTensor:', output)
+
+
+
+
+# Alternatively:
+# Create Nx3 and Nx1 vectors to encode the messages above using BLInputLayer:
+batch=[]
+for batchIdx, msg in enumerate(msgs):
+    l,f=[],[]
+    for y, line in enumerate(msg):
+        for x, c in enumerate(line):
+            if c == 'X':
+                l.append([y, x])  #Locations
+                f.append([1])     #Features
+    batch.append([torch.LongTensor(l),torch.FloatTensor(f)])
+batch=scn.prepare_BLInput(batch)
+batch[1]=batch[1].to(device)
+    
+input = bl_input_layer(batch)
+print('Input SparseConvNetTensor:', input)
+output = model(input)
+
+# Output is 2x32x10x10: our minibatch has 2 samples, the network has 32 output
+# feature planes, and 10x10 is the spatial size of the output.
+print('Output SparseConvNetTensor:', output)
+
+

sparseconvnet/utils.py

@@ -124,14 +124,15 @@ def batch_location_tensors(location_tensors):
             a.append(pad_with_batch_idx(lt,batch_idx))
     return torch.cat(a,0)
 
-def prepare_BLInput(l,f):
+def prepare_BLInput(batch):
     with torch.no_grad():
-        n=max([x.size(0) for x in l])
-        L=torch.empty(len(l),n,l[0].size(1),dtype=torch.int64).fill_(-1)
-        F=torch.zeros(len(l),n,f[0].size(1))
-        for i, (ll, ff) in enumerate(zip(l,f)):
-            L[i,:ll.size(0),:].copy_(ll)
-            F[i,:ff.size(0),:].copy_(ff)
+        n=max([l.size(0) for l,f in batch])
+        l,f=batch[0]
+        L=torch.empty(len(batch),n,l.size(1),dtype=torch.int64).fill_(-1)
+        F=torch.zeros(len(batch),n,f.size(1))
+        for i, (l, f) in enumerate(batch):
+            L[i,:l.size(0),:].copy_(l)
+            F[i,:f.size(0),:].copy_(f)
     return [L,F]
 
 def checkpoint_restore(model,exp_name,name2,use_cuda=True,epoch=0):