initial commit

.gitattributes

@@ -32,3 +32,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+celebrity_custom_model filter=lfs diff=lfs merge=lfs -text
+celebrity_resnet_model filter=lfs diff=lfs merge=lfs -text
+emotion_resnet_model filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,15 @@
+import gradio as gr
+from predict import predict_one_image
+
+input_image = [
+    gr.components.Image(type='filepath',label='Input Image')
+]
+examples = ['joha.jpg','anger.png','bratt.jpg']
+gr.Interface(
+    fn=predict_one_image,
+    inputs=input_image,
+    outputs='text',
+    title="CELEBRITY & EMOTION RECOGNITION APP",
+    examples=examples,
+    cache_examples=False,
+).launch()

celebrity_custom_model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:064745679f779aeb31206e7b7190077367c89dfa97145bc452ab9fb1aafbc1b9
+size 47943319

celebrity_resnet_model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56b67d6619d01cc4fcf6b5639ae59003c3cb878c0e7acaa5a7d95791b07bd154
+size 103103793

emotion_resnet_model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18f37453beaccf41597f85730b803d009d3b109c02d1ccd2144906b01c030e4e
+size 103103793

model.py

@@ -0,0 +1,96 @@
+import torch
+import torch.nn as nn
+
+N_EMOTIONS = 8
+N_CELEBRITIES = 17
+class CustomModel(nn.Module) :
+    def __init__(self,mode = 'emotion') :
+        super().__init__()
+        self.mode = mode
+        
+        self.backbone = nn.Sequential(
+            #3x224x224
+            nn.Conv2d(3, 64, kernel_size=3, stride=1, bias=False),
+            nn.BatchNorm2d(64),
+            nn.LeakyReLU(0.2, inplace=True),
+            # out: 64 x 222 x 222
+
+            nn.Conv2d(64, 32, kernel_size=3, stride=1, bias=False),
+            nn.BatchNorm2d(32),
+            nn.LeakyReLU(0.2, inplace=True),
+            nn.MaxPool2d(kernel_size=2),
+            nn.Dropout(0.2),
+            # out: 32 x 110 x 110
+            
+            nn.Conv2d(32, 32, kernel_size=3, stride=1, bias=False),
+            nn.BatchNorm2d(32),
+            nn.LeakyReLU(0.2, inplace=True),
+            nn.MaxPool2d(kernel_size=2),
+            nn.Dropout(0.3),
+            # out: 32 x 54 x 54
+            
+            
+            
+            
+            nn.Flatten(),
+        )
+        self.in_features = 32*54*54
+        self.neck = nn.Sequential(
+            nn.Linear(self.in_features,128),
+            nn.ReLU(),
+            nn.Linear(128,64),
+            nn.ReLU()
+        )
+        self.emotion_classifier = nn.Linear(64,N_EMOTIONS)
+        self.celebrity_classifier = nn.Linear(64,N_CELEBRITIES)
+    
+    def forward(self,image) :
+        features = self.backbone(image)
+        features = self.neck(features)
+        if self.mode=='emotion' : 
+            emotion_logits = self.emotion_classifier(features)
+            return emotion_logits
+        elif self.mode=='celebrity' :
+            celebrity_logits = self.celebrity_classifier(features)
+            return celebrity_logits
+        else : 
+            emotion_logits = self.emotion_classifier(features)
+            celebrity_logits = self.celebrity_classifier(features)
+            return emotion_logits,celebrity_logits
+
+        
+        
+import torchvision.models as models   
+class ResNet50Model(nn.Module) : 
+    def __init__(self,mode = 'emotion') :
+        super().__init__()
+        self.mode = mode
+        
+        self.backbone = getattr(models, 'resnet50')(False)
+        self.in_features = 1000
+        self.neck = nn.Sequential(
+            nn.Linear(self.in_features,128),
+            nn.ReLU(),
+            nn.Linear(128,64),
+            nn.ReLU()
+        )
+        self.emotion_classifier = nn.Linear(64,N_EMOTIONS)
+        self.celebrity_classifier = nn.Linear(64,N_CELEBRITIES)
+    
+    def forward(self,image) :
+        features = self.backbone(image)
+        features = self.neck(features)
+        if self.mode=='emotion' : 
+            emotion_logits = self.emotion_classifier(features)
+            return emotion_logits
+        elif self.mode=='celebrity' :
+            celebrity_logits = self.celebrity_classifier(features)
+            return celebrity_logits
+        else : 
+            emotion_logits = self.emotion_classifier(features)
+            celebrity_logits = self.celebrity_classifier(features)
+            return emotion_logits,celebrity_logits
+
+
+    
+

predict.py

@@ -0,0 +1,46 @@
+import cv2
+from utils import read_image,get_valid_augs
+import torch
+import torch.nn.functional as F
+from model import ResNet50Model
+import numpy as np
+
+CKPT_EMOTION = 'emotion_resnet_model'
+CKPT_CELEBRITY = 'celebrity_resnet_model'
+FaceInverseTargetMapper = {0: 'Tom Hanks',1: 'Sandra Bullock',2: 'Natalie Portman',3: 'Scarlett Johansson',4: 'Robert Downey Jr',5: 'Nicole Kidman',6: 'Brad Pitt',
+ 7: 'Hugh Jackman',8: 'Tom Cruise',9: 'Leonardo DiCaprio',10: 'Megan Fox',11: 'Johnny Depp',12: 'Will Smith',13: 'Denzel Washington',14: 'Jennifer Lawrence',15: 'Kate Winslet',16: 'Angelina Jolie'}
+EmotionMapper = {0: 'sadness',1: 'contempt',2: 'happiness',3: 'surprise',
+ 4: 'fear',5: 'anger',6: 'disgust',7: 'neutrality'}
+def predict_one_image(path) :
+    image = read_image(path)
+    image = get_valid_augs()(image=image)['image']
+    image = torch.tensor(image,dtype=torch.float)
+    image = image.reshape((1,3,224,224))
+    emotion_model = ResNet50Model('emotion')
+    #loading ckpt
+    emotion_model.load_state_dict(torch.load(CKPT_EMOTION,map_location=torch.device('cpu')))
+
+    celebrity_model = ResNet50Model('celebrity')
+    #loading ckpt
+    celebrity_model.load_state_dict(torch.load(CKPT_CELEBRITY,map_location=torch.device('cpu')))
+    
+    
+    with torch.no_grad() :
+        #emotion
+        outputs = emotion_model(image)
+        outputs = torch.nn.functional.softmax(outputs).cpu().detach().numpy()
+        print(outputs.shape)
+        emotion = np.argmax(outputs,axis=1)[0]
+        emotion_proba = np.max(outputs,axis=1)[0]
+        print(emotion_proba)
+        #celebrity
+        outputs = celebrity_model(image)
+        outputs = torch.nn.functional.softmax(outputs).cpu().detach().numpy()
+        print(outputs.shape)
+        celebrity = np.argmax(outputs,axis=1)[0]
+        celebrity_proba = np.max(outputs,axis=1)[0]
+        print(celebrity_proba)
+    
+    if celebrity_proba<0.45 : 
+        return f"Unkonwn Person Detected with emotion {EmotionMapper[emotion]} "
+    return f"Detected {FaceInverseTargetMapper[celebrity]} with emotion {EmotionMapper[emotion]} "

utils.py

@@ -0,0 +1,22 @@
+import cv2
+import albumentations as A
+from albumentations.pytorch import ToTensorV2
+
+IMAGENET_DEFAULT_MEAN = (0.485, 0.456, 0.406)
+IMAGENET_DEFAULT_STD = (0.229, 0.224, 0.225)
+
+def read_image(path) :
+    img = cv2.imread(path)
+    img = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
+    return img
+
+def get_valid_augs() :
+    return  A.Compose([
+   A.Resize(height=224, width=224, always_apply=True, p=1),
+   A.Normalize(
+        mean = IMAGENET_DEFAULT_MEAN,
+        std  = IMAGENET_DEFAULT_STD,
+        max_pixel_value=255
+            ),
+   ToTensorV2(),
+])