added code and files

.gitignore

@@ -0,0 +1 @@
+/__pycache__/

app.py

@@ -0,0 +1,71 @@
+import os
+import streamlit as st
+from PIL import Image
+
+from inference import get_predictions
+
+
+st.title('Person characteristic prediction Demo')
+
+sample_files = os.listdir('./data/sample_images')
+tot_index = len(sample_files)
+sample_path = './data/sample_images'
+
+if 'image_index' not in st.session_state:
+    st.session_state['image_index'] = 4
+
+if 'which_button' not in st.session_state:
+    st.session_state['which_button'] = 'sample_button'
+
+stream_col, upload_col, sample_col = st.tabs(['Take picture', 'Upload file', 'Select from sample images'])
+with stream_col:
+    picture = st.camera_input("Take a picture")
+    if picture is not None:
+        captured_img = Image.open(picture)
+        st.image(captured_img, caption='Captured Image')
+        use_captured_image = st.button('Use this captured image')
+        if use_captured_image is True:
+            st.session_state['which_button'] = 'captured_button'
+with upload_col:
+    uploaded_file = st.file_uploader("Select a picture from your computer(png/jpg) :", type=['png', 'jpg', 'jpeg'])
+    if uploaded_file is not None:
+        img = Image.open(uploaded_file)
+        st.image(img, caption='Uploaded Image')
+        use_uploaded_image = st.button("Use uploaded image")
+        if use_uploaded_image is True:
+            st.session_state['which_button'] = 'upload_button'
+
+with sample_col:
+    st.write("Select one from these available samples: ")
+    current_index = st.session_state['image_index']
+    current_image = Image.open(os.path.join(sample_path, sample_files[current_index]))
+
+    # next = st.button('next_image')
+    prev_button, next_button = st.columns(2)
+    with prev_button:
+        prev = st.button('prev_image')
+    with next_button:
+        next = st.button('next_image')
+    if prev:
+        current_index = (current_index - 1) % tot_index
+    if next:
+        current_index = (current_index + 1) % tot_index
+    st.session_state['image_index'] = current_index
+    sample_image = Image.open(os.path.join(sample_path, sample_files[current_index]))
+    st.image(sample_image, caption='Chosen image')
+
+    use_sample_image = st.button("Use this Sample")
+    if use_sample_image is True:
+        st.session_state['which_button'] = 'sample_button'
+
+predict_clicked = st.button("Get prediction")
+if predict_clicked:
+    which_button = st.session_state['which_button']
+    if which_button == 'sample_button':
+        predictions = get_predictions(sample_image)
+    elif which_button == 'upload_button':
+        predictions = get_predictions(img)
+    elif which_button  == 'captured_button':
+        predictions = get_predictions(captured_img)
+    st.markdown('**The model predictions along with their probabilities are :**')
+    st.table(predictions)

final-models/resnet_101_weigthed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50caa9aa9f00f30c0ff6ae5e16c487bc3ba3db59ffd57e7010358cd165848252
+size 176824799

inference.py

@@ -0,0 +1,52 @@
+import streamlit as st
+import torch
+import torch.nn.functional as F
+from torchvision.transforms import Compose, Resize, ToTensor, Normalize
+
+from model import AgePredictResnet
+
+path = './final-models/resnet_101_weigthed.pt'
+age_dict = {
+    0: '0 to 10', 1: '10 to 20', 2: '20 to 30', 3: '30 to 40', 4: '40 to 50', 5: '50 to 60',
+    6: '60 to 70', 7: '70 to 80', 8: 'Above 80'
+}
+sex_dict = {0: 'Male', 1: 'Female'}
+race_dict = {
+    0: 'White', 1: 'Black', 2: 'Asian', 3: 'Indian', 4: 'Others (like Hispanic, Latino, Middle Eastern etc)'
+}
+
+@st.experimental_memo
+def load_trained_model(model_path):
+    model = AgePredictResnet()
+    model.load_state_dict(torch.load(model_path, map_location=torch.device('cpu')), strict=False)
+    model.eval()
+    return model
+
+
+def get_predictions(input_image):
+    model = load_trained_model(path)
+    transforms = Compose([Resize((256, 256)), ToTensor(),
+                          Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
+    transformed_image = transforms(input_image)
+    transformed_image = torch.unsqueeze(transformed_image, 0)
+    with torch.inference_mode():
+        logits = model(transformed_image)
+    age_prob = F.softmax(logits[0], dim=1)
+    sex_prob = F.softmax(logits[1], dim=1)
+    race_prob = F.softmax(logits[2], dim=1)
+    top2_age = torch.topk(age_prob, 2, dim=1)
+    sex = torch.argmax(sex_prob, dim=1)
+    top2_race = torch.topk(race_prob, 2, dim=1)
+    all_predictions = (list(top2_age.values.numpy().reshape(-1)), list(top2_age.indices.numpy().reshape(-1))), (
+    sex.item(), sex_prob[0][sex.item()].item()), \
+        (list(top2_race.values.numpy().reshape(-1)), list(top2_race.indices.numpy().reshape(-1)))
+
+    pred_dict = {
+        'Predicted Age range': (age_dict[all_predictions[0][1][0]], age_dict[all_predictions[0][1][1]]),
+        'Age Probability': all_predictions[0][0],
+        'Predicted Sex': sex_dict[all_predictions[1][0]],
+        'Sex Probability': all_predictions[1][1],
+        'Predicted Race': (race_dict[all_predictions[2][1][0]], race_dict[all_predictions[2][1][1]]),
+        'Race Probability': all_predictions[2][0],
+    }
+    return pred_dict

model.py

@@ -0,0 +1,84 @@
+import os
+
+from PIL import Image
+import torch
+import torchvision
+from torch import nn
+import torch.nn.functional as F
+from torchvision.transforms import Compose, Resize, ToTensor, Normalize
+
+
+class AgePredictResnet(nn.Module):
+    def __init__(self):
+        super().__init__()
+        self.model = torchvision.models.resnet101()
+        self.model.fc = nn.Linear(2048, 512)
+        self.age_linear1 = nn.Linear(512, 256)
+        self.age_linear2 = nn.Linear(256, 128)
+        self.age_out = nn.Linear(128, 9)
+        self.gender_linear1 = nn.Linear(512, 256)
+        self.gender_linear2 = nn.Linear(256, 128)
+        self.gender_out = nn.Linear(128, 2)
+        self.race_linear1 = nn.Linear(512, 256)
+        self.race_linear2 = nn.Linear(256, 128)
+        self.race_out = nn.Linear(128, 5)
+        self.activation = nn.ReLU()
+        self.dropout = nn.Dropout(0.4)
+
+    def forward(self, x):
+        out = self.activation(self.model(x))
+        age_out = self.activation(self.dropout((self.age_linear1(out))))
+        age_out = self.activation(self.dropout(self.age_linear2(age_out)))
+        age_out = self.age_out(age_out)
+
+        gender_out = self.activation(self.dropout((self.gender_linear1(out))))
+        gender_out = self.activation(self.dropout(self.gender_linear2(gender_out)))
+        gender_out = self.gender_out(gender_out)
+
+        race_out = self.activation(self.dropout((self.race_linear1(out))))
+        race_out = self.activation(self.dropout(self.race_linear2(race_out)))
+        race_out = self.race_out(race_out)
+        return age_out, gender_out, race_out
+
+
+if __name__ == '__main__':
+    trained_model_path = os.path.join('./final-models/resnet_101_weigthed.pt')
+    model = AgePredictResnet()
+    model.load_state_dict(torch.load(trained_model_path, map_location=torch.device('cpu')), strict=False)
+    model.eval()
+    sample_image = Image.open('../../age_prediction/data/wild_images/part1/50_1_1_20170110120147003.jpg')
+    transforms = Compose([Resize((256, 256)), ToTensor(),
+                          Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
+    transformed_image = transforms(sample_image)
+    transformed_image = torch.unsqueeze(transformed_image, 0)
+    print(transformed_image.shape)
+    with torch.inference_mode():
+        logits = model(transformed_image)
+    age_prob = F.softmax(logits[0], dim=1)
+    sex_prob = F.softmax(logits[1], dim=1)
+    race_prob = F.softmax(logits[2], dim=1)
+    top2_age = torch.topk(age_prob, 2, dim=1)
+    sex = torch.argmax(sex_prob, dim=1)
+    top2_race = torch.topk(race_prob, 2, dim=1)
+    all_predictions = (list(top2_age.values.numpy().reshape(-1)), list(top2_age.indices.numpy().reshape(-1))), (
+    sex.item(), sex_prob[0][sex.item()].item()), \
+        (list(top2_race.values.numpy().reshape(-1)), list(top2_race.indices.numpy().reshape(-1)))
+    print(all_predictions)
+    age_dict = {
+        0: '0 to 10', 1: '10 to 20', 2: '20 to 30', 3: '30 to 40', 4: '40 to 50', 5: '50 to 60',
+        6: '60 to 70', 7: '70 to 80', 8: 'Above 80'
+    }
+    sex_dict = {0: 'Male', 1: 'Female'}
+    race_dict = {
+        0: 'White', 1: 'Black', 2: 'Asian', 3: 'Indian', 4: 'Others (like Hispanic, Latino, Middle Eastern etc)'
+    }
+    #
+    pred_dict = {
+        'Predicted Age range': (age_dict[all_predictions[0][1][0]], age_dict[all_predictions[0][1][1]]),
+        'Age Probability': all_predictions[0][0],
+        'Predicted Sex': sex_dict[all_predictions[1][0]],
+        'Sex Probability': all_predictions[1][1],
+        'Predicted Race': (race_dict[all_predictions[2][1][0]], race_dict[all_predictions[2][1][1]]),
+        'Race Probability': all_predictions[2][0],
+    }
+    print(pred_dict)

requirements.txt

@@ -0,0 +1,53 @@
+--extra-index-url https://download.pytorch.org/whl/cpu
+altair==4.2.0
+attrs==22.2.0
+backports.zoneinfo==0.2.1
+blinker==1.5
+cachetools==5.2.0
+certifi==2022.12.7
+charset-normalizer==2.1.1
+click==8.1.3
+commonmark==0.9.1
+decorator==5.1.1
+entrypoints==0.4
+gitdb==4.0.10
+GitPython==3.1.29
+idna==3.4
+importlib-metadata==5.2.0
+importlib-resources==5.10.1
+Jinja2==3.1.2
+jsonschema==4.17.3
+MarkupSafe==2.1.1
+numpy==1.24.0
+packaging==22.0
+pandas==1.5.2
+Pillow==9.3.0
+pkgutil_resolve_name==1.3.10
+protobuf==3.20.3
+pyarrow==10.0.1
+pydeck==0.8.0
+Pygments==2.13.0
+Pympler==1.0.1
+pyrsistent==0.19.2
+python-dateutil==2.8.2
+pytz==2022.7
+pytz-deprecation-shim==0.1.0.post0
+requests==2.28.1
+rich==12.6.0
+semver==2.13.0
+six==1.16.0
+smmap==5.0.0
+streamlit==1.16.0
+toml==0.10.2
+toolz==0.12.0
+torch==1.13.1
+torchaudio==0.13.1
+torchvision==0.14.1
+tornado==6.2
+typing_extensions==4.4.0
+tzdata==2022.7
+tzlocal==4.2
+urllib3==1.26.13
+validators==0.20.0
+watchdog==2.2.0
+zipp==3.11.0