fix

app/Hackathon_setup/exp_recognition.py

@@ -8,11 +8,8 @@ from PIL import Image
 import base64
 import io
 import os
-import torchvision.models as models
 ## Add more imports if required
 
-classes = {0: 'ANGER', 1: 'DISGUST', 2: 'FEAR', 3: 'HAPPINESS', 4: 'NEUTRAL', 5: 'SADNESS', 6: 'SURPRISE'}
-
 #############################################################################################################################
 #   Caution: Don't change any of the filenames, function names and definitions                                              #
 #   Always use the current_path + file_name for refering any files, without it we cannot access files on the server         #
@@ -51,40 +48,40 @@ def detected_face(image):
 #4) Perform necessary transformations to the input(detected face using the above function), this should return the Expression in string form ex: "Anger"
 #5) For loading your model use the current_path+'your model file name', anyhow detailed example is given in comments to the function 
 ##Caution: Don't change the definition or function name; for loading the model use the current_path for path example is given in comments to the function
-def get_expression(img):
+def get_expression(img_str):
     device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-    
+    imgdata = base64.b64decode(img_str)
+    img = Image.open(io.BytesIO(imgdata))
+    img =  np.array(img.getdata()).reshape(img.size[1], img.size[0], 3).astype(np.uint8)
+
     ##########################################################################################
     ##Example for loading a model using weight state dictionary:                            ##
-    ## face_det_net = facExpRec() #Example Network                                          ##
-    ## model = torch.load(current_path + '/exp_recognition_net.t7', map_location=device)    ##
-    ## face_det_net.load_state_dict(model['net_dict'])                                      ##
+    face_det_net = Model().to(device) #Example Network                                          ##
+    model = torch.load(current_path + '/expression_model.t7', map_location=device)    ##
+    face_det_net.load_state_dict(model['net_dict'])
+    #face_det_net.to(device)
+    #face_det_net.eval()                                      ##
     ##                                                                                      ##
     ##current_path + '/<network_definition>' is path of the saved model if present in       ##
     ##the same path as this file, we recommend to put in the same directory                 ##
     ##########################################################################################
     ##########################################################################################
-    model_path = current_path + '/expression_model.t7'
-    loaded_model = models.resnet18(pretrained=False)
-    num_ftrs = loaded_model.fc.in_features
-    loaded_model.fc = nn.Linear(num_ftrs, 7)
-
-    loaded_model.load_state_dict(torch.load(model_path, map_location='cpu')['net_dict'])
-    loaded_model.eval()
-
+    
     face = detected_face(img)
     if face==0:
-        face = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2GRAY))
+        return "No Face Found"
+    #Apply transformation to the image to convert to a tensor that can be passed to your model
+    trnscm = transforms.Compose([rgb2gray(face), transforms.Resize((100,100)), transforms.ToTensor()])
     
-    face1 = trnscm(face).unsqueeze(0)
-    # YOUR CODE HERE, return expression using your model
-
+    #Call the forward function of your model to get the prediction on the transformed image
     with torch.no_grad():
-        output = loaded_model(face1)
-        output = output.cpu()
-        _, predicted = torch.max(output, 1)
-
-
-    predicted_label = classes[predicted.item()]
+        features = face_det_net.forward(trnscm)
+    #Take the argmax to get the index of the predicted class
+    predicted_class = np.argmax(features)
+    
+    #classes = ['Anger', 'person2', 'person3', 'person4', 'person5', 'person6', 'person7']
+    #From the class index predicted, return the corresponding expression string
+    string = classes[predicted_class]
+    # YOUR CODE HERE, return expression using your model
 
-    return predicted_label
+    return string

app/Hackathon_setup/exp_recognition_model.py

@@ -2,6 +2,7 @@ import torch
 import torchvision
 import torch.nn as nn
 from torchvision import transforms
+import torch.nn.functional as F
 ## Add more imports if required
 
 ####################################################################################################################
@@ -13,18 +14,50 @@ from torchvision import transforms
 classes = {0: 'ANGER', 1: 'DISGUST', 2: 'FEAR', 3: 'HAPPINESS', 4: 'NEUTRAL', 5: 'SADNESS', 6: 'SURPRISE'}
 
 # Example Network
-class facExpRec(torch.nn.Module):
+class Model(nn.Module):
     def __init__(self):
-        pass   # remove 'pass' once you have written your code
-        #YOUR CODE HERE
+        super(Model, self).__init__()
         
+       
+        self.conv1 = nn.Conv2d(in_channels=1, out_channels=16, kernel_size=3)
+         
+                   
+        self.conv2 = nn.Conv2d(in_channels=16, out_channels=64, kernel_size=3)
+
+        self.conv3 = nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3)
+        
+        # Define the Fully connected layers
+        # The output of the second convolution layer will be input to the first fully connected layer
+        self.fc1 = nn.Linear(128*10*10, 256)
+        # 256 input features, 128 output features 
+        self.fc2 = nn.Linear(256, 128)
+        # 128 input features, 64 output features 
+        self.fc3 = nn.Linear(128, 64)
+        # 64 input features, 7 output features for our 7 defined classes
+        self.fc4 = nn.Linear(64, 7)
+
+        # Max pooling
+        self.pool = nn.MaxPool2d(kernel_size=2)  # Max pooling layer with filter size 2x2
+   
     def forward(self, x):
-        pass   # remove 'pass' once you have written your code
-        #YOUR CODE HERE
+
+        x = self.pool(F.relu(self.conv1(x)))
+        x = self.pool(F.relu(self.conv2(x)))
+        x = self.pool(F.relu(self.conv3(x)))
+        # Flatten the image
+        x = x.view(-1, 128*10*10)  # Output shape of convolutional layer is 16*5*5 
+        
+        # Linear layers with RELU activation
+        x = F.relu(self.fc1(x))
+        x = F.relu(self.fc2(x))
+        x = F.relu(self.fc3(x))
+        x = self.fc4(x)
+        x = F.log_softmax(x, dim=1)
+        return x
         
 # Sample Helper function
-# def rgb2gray(image):
-#     return image.convert('L')
+def rgb2gray(image):
+    return image.convert('L')
     
 # Sample Transformation function
 #YOUR CODE HERE for changing the Transformation values.

app/Hackathon_setup/expression_model.t7

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a06b848d5b3aec70945a05fcebd7a113ece8615efe43b681f56d2b81125b777b
-size 94410006
+oid sha256:f1fb4914225fd633bc6391ee3808bce815165c9fe8822e8ffe5f5e172adfb942
+size 13612942