Merge pull request CSAILVision#53 from CSAILVision/hang-dev

 minor re-orgnization

Author: hangzhaomit

.gitignore

@@ -1,4 +1,3 @@
-__init__.py
 *.pyc
 
 ckpt/

README.md

@@ -94,7 +94,7 @@ IMPORTANT: We use our self-trained base model on ImageNet. The model takes the i
     <tr>
         <td rowspan="2"><b>UperNet-101</b></td>
         <td>No</td><td>41.98</td><td>80.63</td><td>61.34</td>
-        <td rowspan="2">2.5 * 25 = 50.0 hours</td>
+        <td rowspan="2">2.5 * 25 = 62.5 hours</td>
     </tr>
     <tr>
         <td>Yes</td><td>42.66</td><td>81.01</td><td>61.84</td>
@@ -115,11 +115,6 @@ The code is developed under the following configurations.
 
 *Warning:* We don't support the outdated Python 2 anymore. PyTorch 0.4.0 or higher is required to run the codes.
 
-As PyTorch 0.4.0 is not officially released yet. We suggest running in a separate conda environment, build PyTorch from source, or install the following PyTorch package:
-```bash
-conda install -c ostrokach-forge pytorch=0.4.0
-```
-
 ## Quick start: Test on an image using our trained model 
 1. Here is a simple demo to do inference on a single image:
 ```bash

models/__init__.py

@@ -0,0 +1 @@
+from .models import ModelBuilder, SegmentationModule

models.py renamed to models/models.py

@@ -1,8 +1,7 @@
 import torch
 import torch.nn as nn
 import torchvision
-import resnet
-import resnext
+from . import resnet, resnext
 from lib.nn import SynchronizedBatchNorm2d
 
 
@@ -531,4 +530,3 @@ def forward(self, conv_out, segSize=None):
         x = nn.functional.log_softmax(x, dim=1)
 
         return x
-