big update changet tranfrorms

Author: Sylwester Dawida

.gitignore

@@ -9,4 +9,5 @@ python/data
 python/kinect_leap_dataset
 python/openCvTranforms
 python/laRED_dataset
-kivy_venv
+kivy_venv
+doc

python/.idea/workspace.xml

@@ -114,21 +114,23 @@ def __getitem__(self, idx):
     split = 2
     import openCvTranforms.opencv_transforms.transforms as tf
 
-    heigh = 60
-    width = 60
+    heigh = 200
+    width = 200
 
 
     def imshow(img):
-        # img = img.view(3, heigh, width, 1). / 2 + 0.5  # unnormalize
+        img = img[0].permute(1, 2, 0).numpy() / 2 + 0.5
+        # img = img.view(heigh, width, 3).numpy()   # unnormalize
         print(img.shape)
-        # npimg = img.numpy()
-        cv2.imshow("preview", img.numpy())
+        cv2.imshow("preview", img)
 
 
     transform = transforms.Compose(
-        [tf.Resize((heigh, width)),
+        [tf.RandomCrop([400, 400]),
+         tf.Resize((heigh, width)),
+         tf.ColorJitter(brightness=0.5, contrast=0.5),
          tf.ToTensor(),
-         tf.Normalize((0.5,), (0.5,))
+         tf.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
          ])
     dir = "D:/DataSetNew/"
     testDataset = myDataset(dir, split, test=True, transform=transform)
@@ -145,5 +147,10 @@ def imshow(img):
                                              num_workers=4)
 
     trainLoader = torch.utils.data.DataLoader(trainDataset,
-                                              batch_size=1, shuffle=True,
+                                              batch_size=1, shuffle=False,
                                               num_workers=4)
+
+    for data in trainLoader:
+        images, labels = data
+        imshow(images)
+        cv2.waitKey(0)

python/__pycache__/datasetLoader.cpython-37.pyc

@@ -6,6 +6,7 @@
 from time import time
 from enum import Enum
 from comunication import Commands
+
 if __name__ == '__main__':
 
     device = cv2.VideoCapture(0, cv2.CAP_DSHOW)
@@ -62,6 +63,7 @@ class pearsons(Enum):
         AD = "AD"
         KH = "KH"
 
+
     """        
         default = 1
         turnLeft = 2
@@ -81,8 +83,8 @@ class pearsons(Enum):
     add = True
     # if false all files in existing dir will be deleted
     saving = True
-    command = commands.stopCommand
-    pearson = pearsons.JD
+    command = commands.slowDown
+    pearson = pearsons.KH
     howMany = 100
     savePerSec = 12.0
     ##end settigns
@@ -109,7 +111,7 @@ class pearsons(Enum):
         print("Old files deleted")
     print("files: {}".format(file))
     print("Creating dataset. Gesture: {}. Pearson: {}".format(command.name, pearson.name))
-    check = cv2.imread("D:\DataSetNew\{}\SD\{}_1.jpg".format(command.name,command.value))
+    check = cv2.imread("D:\DataSetNew\{}\SD\{}_1.jpg".format(command.name, command.value))
     frameWithStats = cv2.flip(check, 1)
     frameWithStats = drawStatistics(frameWithStats, out, 1)
     cv2.imshow("AllImage", frameWithStats)

python/__pycache__/main.cpython-37.pyc

@@ -52,7 +52,7 @@ def __init__(self, print):
         self.transform = transforms.Compose([
             tf.Resize((self.heigh, self.width)),
             tf.ToTensor(),
-            tf.Normalize((0.5,), (0.5,))
+            tf.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
         ])
         self.signals = Signals()
         self.running = True

python/__pycache__/mainCuda.cpython-37.pyc

@@ -9,6 +9,7 @@
 import openCvTranforms.opencv_transforms.transforms as tf
 import torchvision.transforms as transforms
 import keyboard
+from resultsPresentation import updating
 
 
 # plot image with opencv
@@ -54,6 +55,8 @@ def train(net, epochs, startingEpoch):
     if startEpoch > epochs:
         return epochs + 1
 
+    plot = updating()
+
     with open(logFile, "a") as f:
         for epoch in range(startingEpoch, EPOCHS):
             print("\aepoch: " + str(epoch))
@@ -73,11 +76,12 @@ def train(net, epochs, startingEpoch):
 
                 acc, loss = fwd_pass(images, labels, train=True)
 
-                if i % 80 == 79:
-                    val_acc, val_loss = test(64)
+                if i % 100 == 99:
+                    val_acc, val_loss = test()
                     # print(val_acc, float(val_loss))
                     f.write(
                         f"{MODEL_NAME},{round(time.time(), 3)},{round(float(acc), 3)}, {round(float(loss), 4)}, {round(float(val_acc), 3)}, {round(float(val_loss), 4)}, {epoch}\n")
+                    plot.update(acc, loss, val_acc, val_loss)
                 i += 1
             t.close()
             if exit:
@@ -86,7 +90,7 @@ def train(net, epochs, startingEpoch):
 
 
 class globNr(object):
-    nr = 19
+    nr = 21
 
 
 if __name__ == "__main__":
@@ -103,21 +107,30 @@ class globNr(object):
     modelFilename = "./savedMode{}.pth".format(nr)
     dataSetPath = "D:/DataSetNew/"
     logFile = 'model{}.log'.format(nr)
-    loadTrainData = True
+    loadTrainData = False
     split = 1
     width = 60
     heigh = 60
     epochs = 10
-    batchSize = 15
+    batchSize = 30
     startEpoch = 0
-
-    transform = transforms.Compose(
-        [tf.Resize((heigh, width)),
-         tf.ToTensor(),
-         tf.Normalize((0.5,), (0.5,))])
-
+    drawing = False
+
+    transform = transforms.Compose([
+        tf.Resize((heigh, width)),
+        tf.ToTensor(),
+        tf.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
+    ])
+
+    trainTransform = transforms.Compose([
+        tf.RandomCrop([400, 400]),
+        tf.Resize((heigh, width)),
+        tf.ColorJitter(brightness=0.5, contrast=0.5),
+        tf.ToTensor(),
+        tf.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
+    ])
     testLoader = myDataset(dataSetPath, split, test=True, transform=transform)
-    trainLoader = myDataset(dataSetPath, split, test=False, transform=transform)
+    trainLoader = myDataset(dataSetPath, split, test=False, transform=trainTransform)
     print("test {}, train {}".format(len(testLoader), len(trainLoader)))
     testDataset = torch.utils.data.DataLoader(testLoader,
                                               batch_size=batchSize, shuffle=True)

python/__pycache__/neuralNetwork.cpython-37.pyc

@@ -0,0 +1,44 @@
+model-1578229518,1578229670.18,0.1, 0.0709, 0.13, 0.0717, 0
+model-1578229518,1578229799.806,0.133, 0.0707, 0.166, 0.0698, 0
+model-1578229518,1578229918.962,0.3, 0.0647, 0.185, 0.0689, 0
+model-1578229518,1578230022.535,0.167, 0.0708, 0.168, 0.0689, 0
+model-1578229518,1578230125.936,0.233, 0.062, 0.222, 0.0658, 0
+model-1578229518,1578230225.786,0.267, 0.0616, 0.215, 0.0651, 0
+model-1578229518,1578230321.838,0.133, 0.0671, 0.202, 0.0653, 0
+model-1578229518,1578230415.446,0.2, 0.0642, 0.204, 0.0655, 0
+model-1578229518,1578230501.56,0.267, 0.0662, 0.214, 0.0642, 0
+model-1578229518,1578230583.264,0.167, 0.0635, 0.236, 0.0623, 0
+model-1578229518,1578230663.222,0.133, 0.0637, 0.232, 0.063, 0
+model-1578229518,1578230726.592,0.167, 0.0641, 0.229, 0.0627, 1
+model-1578229518,1578230770.25,0.2, 0.0604, 0.234, 0.062, 1
+model-1578229518,1578230813.477,0.2, 0.0589, 0.24, 0.0621, 1
+model-1578229518,1578230857.516,0.133, 0.0664, 0.228, 0.0618, 1
+model-1578229518,1578230900.095,0.333, 0.0612, 0.239, 0.0618, 1
+model-1578229518,1578230945.273,0.133, 0.0679, 0.234, 0.0614, 1
+model-1578229518,1578230995.203,0.333, 0.0576, 0.258, 0.0615, 1
+model-1578229518,1578231048.311,0.267, 0.0576, 0.256, 0.0612, 1
+model-1578229518,1578231106.379,0.233, 0.0656, 0.278, 0.0608, 1
+model-1578229518,1578231174.712,0.2, 0.0639, 0.279, 0.0601, 1
+model-1578229518,1578231244.272,0.233, 0.0654, 0.289, 0.0591, 1
+model-1578231672,1578231722.662,0.133, 0.0662, 0.281, 0.0659, 0
+model-1578231672,1578231769.662,0.3, 0.0627, 0.441, 0.0535, 0
+model-1578231672,1578231821.712,0.467, 0.0557, 0.539, 0.0485, 0
+model-1578231672,1578231876.484,0.533, 0.0473, 0.632, 0.0404, 0
+model-1578231672,1578231944.577,0.667, 0.0415, 0.646, 0.0394, 0
+model-1578231672,1578232022.542,0.567, 0.0436, 0.744, 0.0311, 0
+model-1578231672,1578232107.579,0.633, 0.0344, 0.799, 0.0263, 0
+model-1578231672,1578232183.972,0.733, 0.0262, 0.808, 0.0238, 1
+model-1578231672,1578232217.037,0.633, 0.0343, 0.842, 0.0216, 1
+model-1578231672,1578232249.613,0.767, 0.0321, 0.869, 0.0192, 1
+model-1578231672,1578232283.161,0.7, 0.0338, 0.844, 0.0212, 1
+model-1578231672,1578232322.598,0.733, 0.0293, 0.853, 0.0192, 1
+model-1578231672,1578232370.007,0.733, 0.0263, 0.857, 0.0197, 1
+model-1578231672,1578232436.885,0.933, 0.0115, 0.903, 0.0157, 1
+model-1578231672,1578232512.7,0.9, 0.0156, 0.881, 0.017, 2
+model-1578231672,1578232548.062,0.833, 0.0223, 0.875, 0.0181, 2
+model-1578231672,1578232583.239,0.833, 0.0171, 0.891, 0.0152, 2
+model-1578231672,1578232620.201,0.833, 0.0184, 0.907, 0.0136, 2
+model-1578231672,1578232658.165,0.833, 0.0255, 0.905, 0.0141, 2
+model-1578231672,1578232701.094,0.8, 0.026, 0.898, 0.015, 2
+model-1578231672,1578232757.489,0.733, 0.0291, 0.921, 0.0131, 2
+model-1578233527,1578233556.541,0.067, 0.0709, 0.077, 0.0733, 0