Merge pull request #4 from MalondaClement/test_fix_tunnel

Fix tunnel dataset

README.md

@@ -9,7 +9,7 @@ You can read the documentation [here](https://github.com/MalondaClement/pipeline
 
 ## 2. How to use the pipeline
 
-```python
+```bash
 git clone https://github.com/MalondaClement/pipeline.git
 ```
 

datasets/tunnel.py

@@ -43,7 +43,7 @@ def __init__(self, root, split="train", labels_type="csv", transform=None, targe
         assert split in ["train","val","test"], "Unknown value {} for argument split.".format(split)
         if self.labels_type == "json":
             self.target_dir = os.path.join(self.root, "jsons")
-            self.__read_csv()
+            self.__read_json()
         elif self.labels_type == "csv":
             self.target_dir = os.path.join(self.root, "csvs")
             self.__read_csv()
@@ -62,6 +62,8 @@ def __getitem__(self, index):
             draw.polygon(e[0], fill=classToVal[e[1]])
         image = np.array(image)
         image = image.transpose(2, 0, 1)
+        # int to float to fix training
+        image = image/255
         target = np.array(target)[:, :, 0]
         return image, target, filepath
 

inference.py

@@ -24,7 +24,7 @@ def main():
     args = ARGS("DeepLabV3_Resnet50", "tunnel", len(Dataset.validClasses), labels_type="csv", batch_size=2, epochs=2)
 
     model, args = get_model(args)
-    args.save_path = "path"
+    args.save_path = "../"
 
     checkpoint = torch.load(os.path.join(args.save_path, "best_weights.pth.tar"), map_location=torch.device('cpu'))
 
@@ -44,10 +44,12 @@ def main():
     times = list()
 
     start = time.time()
-    for i, file in enumerate(os.listdir("images_inf_path")):
+    for i, file in enumerate(os.listdir("/Users/ClementMalonda/Desktop/img_inf")):
+        if file[-4:] != ".png":
+            continue
         start = time.time()
 
-        img = Image.open(os.path.join("images_inf_path",file))
+        img = Image.open(os.path.join("/Users/ClementMalonda/Desktop/img_inf",file))
         img = np.array(img)
         img = img[:,:,:3]
         img = img/255

learning/learner.py

@@ -41,6 +41,19 @@ def train_epoch(dataloader, model, criterion, optimizer, lr_scheduler, epoch, va
         # Iterate over data.
         for epoch_step, (inputs, labels, _) in enumerate(dataloader):
             data_time.update(time.time()-end)
+
+            #test
+            fig, (ax0, ax1, ax2) = plt.subplots(1, 3)
+            images = inputs.numpy()
+            image = images[0, :, :, :]
+            image = image.transpose(1, 2, 0)
+            ax0.imshow(image)
+            ax2.imshow(image)
+            ax0.set_title("Image d'origine")
+            print(type(inputs))
+            print(inputs.shape)
+            #end test
+
             if args.copyblob:
                 for i in range(inputs.size()[0]):
                     rand_idx = np.random.randint(inputs.size()[0])
@@ -72,6 +85,20 @@ def train_epoch(dataloader, model, criterion, optimizer, lr_scheduler, epoch, va
             if args.is_pytorch_model :
                 outputs = outputs['out'] #FIXME for DeepLab V3
             preds = torch.argmax(outputs, 1)
+
+            #test
+            print(type(preds))
+            print(preds.shape)
+            pred = preds[0, :, :].cpu()
+            ax1.imshow(pred)
+            ax1.set_title("Prédiction")
+            ax2.imshow(pred, alpha=0.5)
+            ax2.set_title("Superposition de l'image avec la prédiction")
+            if not os.path.isdir(os.path.join(args.save_path, "inference")):
+                os.makedirs(os.path.join(args.save_path, "inference"))
+            fig.savefig(os.path.join(args.save_path, "inference", str(epoch)+".png"))
+            # end test
+
             # cross-entropy loss
             loss = criterion(outputs, labels)
 

train.py

@@ -14,15 +14,15 @@
 from helpers.helpers import plot_learning_curves
 from learning.learner import train_epoch, validate_epoch
 from learning.utils import get_dataloader
-# from datasets.tunnel import Tunnel
-from datasets.minicity import MiniCity
+from datasets.tunnel import Tunnel
+# from datasets.minicity import MiniCity
 
 def main():
     # Get tunnel dataset
-    Dataset = MiniCity
+    Dataset = Tunnel
 
     # Set up execution arguments
-    args = ARGS("DeepLabV3_MobileNetV3", "microcity", len(Dataset.validClasses), labels_type="csv", batch_size=2, epochs=40)
+    args = ARGS("DeepLabV3_Resnet50", "tunnel", len(Dataset.validClasses), labels_type="json", batch_size=4, epochs=100)
 
     # Get model
     model, args = get_model(args)