diff --git a/Accuracy.png b/Accuracy.png
new file mode 100644
index 0000000..be6e78e
Binary files /dev/null and b/Accuracy.png differ
diff --git a/data_batch_1 b/data_batch_1
new file mode 100644
index 0000000..ab404a5
Binary files /dev/null and b/data_batch_1 differ
diff --git a/data_batch_2 b/data_batch_2
new file mode 100644
index 0000000..6bf1369
Binary files /dev/null and b/data_batch_2 differ
diff --git a/data_batch_3 b/data_batch_3
new file mode 100644
index 0000000..66a0d63
Binary files /dev/null and b/data_batch_3 differ
diff --git a/data_batch_4 b/data_batch_4
new file mode 100644
index 0000000..cf8d03d
Binary files /dev/null and b/data_batch_4 differ
diff --git a/data_batch_5 b/data_batch_5
new file mode 100644
index 0000000..468b2aa
Binary files /dev/null and b/data_batch_5 differ
diff --git a/graph.png b/graph.png
new file mode 100644
index 0000000..b3b3df0
Binary files /dev/null and b/graph.png differ
diff --git a/main.py b/main.py
index dfe56fe..3d57d89 100644
--- a/main.py
+++ b/main.py
@@ -12,6 +12,116 @@
 
 # Import whatever libraries/modules you need
 
+from cProfile import label
+from cgi import test
+from unittest import result
+import matplotlib.pyplot as plt
 import numpy as np
+import PIL
+import pandas as pd
+import torch
+from torch import nn
+import torchvision
+import torchvision.transforms as transforms
+import torch.optim as optim
+import matplotlib.pyplot as plt
 
 # Your working code here
+
+# Unpack data
+def unpickle(file):
+  import pickle
+  with open(file, 'rb') as fo:
+      dict = pickle.load(fo, encoding='bytes')
+  return dict
+
+# Put data into dataframe
+dicts = [];
+for i in range(1,6):
+  dicts.append(unpickle("data_batch_"+str(i)))
+dataFrame = pd.DataFrame(columns=['labels','data'])
+for i in range(5):
+  temp = pd.DataFrame([dicts[i][b'labels'],dicts[i][b'data']])
+  temp = temp.transpose()
+  temp.columns=['labels','data']
+  dataFrame = pd.concat([dataFrame,temp], ignore_index = True)
+
+# Change data into tensor form
+labelTensor = torch.tensor(dataFrame['labels'].values, dtype = torch.uint8) 
+dataTensor = torch.tensor(dataFrame['data'].values, dtype = torch.uint8)
+dataTensor = torch.div(dataTensor,255)
+print(dataTensor)
+
+# Construct neural network
+class NeuralNetwork(nn.Module):
+  def __init__(self):
+    super(NeuralNetwork, self).__init__()
+    self.conv1 = nn.Conv2d(3, 3, 3)
+    self.Flatten = nn.Flatten()
+    self.pool = nn.MaxPool2d(2, 1)
+    self.conv2 = nn.Conv2d(3, 1, 3)
+    self.linearReLUStack = nn.Sequential(
+      nn.Flatten(),
+      nn.Linear(22*22,512),
+      nn.ReLU(),
+    )
+  def forward(self,x):
+    logits = torch.reshape(x, (32,32,3))
+    logits = logits.permute(2,0,1)
+    logits = self.conv1(logits)
+    logits = self.conv1(logits)
+    logits = self.pool(logits)
+    logits = self.conv1(logits)
+    logits = self.conv2(logits)
+    logits = self.pool(logits)
+    logits = self.linearReLUStack(logits)
+    logits = torch.squeeze(logits)
+    return logits
+network = NeuralNetwork()
+loss_fn = torch.nn.CrossEntropyLoss()
+criterion = nn.CrossEntropyLoss()
+optimizer = optim.SGD(network.parameters(), lr = 0.0001, momentum = 0.9)
+
+# Train
+trainLoss = []
+valLoss = []
+for epoch in range(16):
+  print('Epoch ' + str(epoch))
+  runningLoss = 0.0
+  runningValLoss = 0.0
+  for i in range(40000):
+    # 40000
+    optimizer.zero_grad()
+    outputs = network(dataTensor[i])
+    # print(outputs.size())
+    loss = criterion(outputs, labelTensor[i])
+    loss.backward()
+    optimizer.step()
+    runningLoss += loss.item()
+    if(i%2000 == 1999):
+      print("Running loss: " + str(runningLoss/(i+1)))
+  runningLoss = runningLoss/40000
+  trainLoss.append(runningLoss)
+  for i in range(10000):
+    # 10000
+    valOut = network(dataTensor[i+40000])
+    loss = criterion(valOut,labelTensor[i+40000])
+    runningValLoss += loss.item()
+  runningValLoss = runningValLoss/10000
+  valLoss.append(runningValLoss)
+print(trainLoss)
+print(valLoss)
+
+correct = 0
+for i in range(10000):
+  result = network(dataTensor[i+40000])
+  print(torch.argmax(result).item())
+  print(labelTensor.data[i+40000].item())
+  if(torch.argmax(result).item() == labelTensor.data[i+40000].item()):
+    correct += 1
+
+print('Correct: ' + str(correct) + ' (' + str(correct/100.0) + '%) out of 10000')
+plt.plot(trainLoss, label = "Training Loss")
+plt.plot(valLoss, label = "Value Loss")
+plt.legend()
+plt.show()
\ No newline at end of file