Upload 3 files

README.md

@@ -1,12 +1,29 @@
 ---
-title: Violence Detection App
-emoji: 🏢
-colorFrom: purple
+title: Violence Detection 3D CNN
+emoji: 🎥
+colorFrom: red
 colorTo: yellow
-sdk: gradio
-sdk_version: 6.5.1
+sdk: streamlit
+sdk_version: 1.28.0
 app_file: app.py
 pinned: false
+license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# 🎥 Violence Detection in Videos
+
+A Streamlit application that uses a **3D CNN** model to detect violence in video clips.
+
+## Model
+- **Architecture:** 3D CNN (4 convolutional blocks)
+- **Dataset:** RWF-2000 (Real World Fighting)
+- **Task:** Binary Classification (Violence vs Non-Violence)
+- **Input:** 16 frames × 112 × 112 RGB
+
+## How to Use
+1. Upload a video file (AVI, MP4, or MKV)
+2. Click "Analyze Video"
+3. View the prediction results
+
+## Model Card
+[https://huggingface.co/santa47/violence-detection-3dcnn](https://huggingface.co/santa47/violence-detection-3dcnn)

app (2).py

@@ -0,0 +1,235 @@
+import streamlit as st
+import torch
+import torch.nn as nn
+import numpy as np
+import cv2
+import tempfile
+import os
+from huggingface_hub import hf_hub_download
+
+# ============================================
+# MODEL DEFINITION
+# ============================================
+class Conv3DBlock(nn.Module):
+    def __init__(self, in_ch, out_ch, kernel=3, stride=1, padding=1):
+        super().__init__()
+        self.conv = nn.Conv3d(in_ch, out_ch, kernel, stride, padding)
+        self.bn = nn.BatchNorm3d(out_ch)
+        self.relu = nn.ReLU(inplace=True)
+    
+    def forward(self, x):
+        return self.relu(self.bn(self.conv(x)))
+
+
+class ViolenceDetector3DCNN(nn.Module):
+    def __init__(self, num_classes=2, dropout=0.5):
+        super().__init__()
+        
+        self.features = nn.Sequential(
+            Conv3DBlock(3, 64),
+            nn.MaxPool3d((1, 2, 2), (1, 2, 2)),
+            
+            Conv3DBlock(64, 128),
+            nn.MaxPool3d((2, 2, 2), (2, 2, 2)),
+            
+            Conv3DBlock(128, 256),
+            Conv3DBlock(256, 256),
+            nn.MaxPool3d((2, 2, 2), (2, 2, 2)),
+            
+            Conv3DBlock(256, 512),
+            Conv3DBlock(512, 512),
+            nn.MaxPool3d((2, 2, 2), (2, 2, 2)),
+        )
+        
+        self.gap = nn.AdaptiveAvgPool3d((1, 1, 1))
+        
+        self.classifier = nn.Sequential(
+            nn.Dropout(dropout),
+            nn.Linear(512, 256),
+            nn.ReLU(inplace=True),
+            nn.Dropout(dropout),
+            nn.Linear(256, num_classes)
+        )
+    
+    def forward(self, x):
+        x = self.features(x)
+        x = self.gap(x)
+        x = x.view(x.size(0), -1)
+        x = self.classifier(x)
+        return x
+
+
+# ============================================
+# LOAD MODEL
+# ============================================
+@st.cache_resource
+def load_model():
+    # Download model from Hugging Face
+    model_path = hf_hub_download(
+        repo_id="santa47/violence-detection-3dcnn",
+        filename="violence_detector.pth"
+    )
+    
+    # Load model
+    model = ViolenceDetector3DCNN(num_classes=2, dropout=0.5)
+    checkpoint = torch.load(model_path, map_location=torch.device('cpu'))
+    model.load_state_dict(checkpoint['model_state_dict'])
+    model.eval()
+    
+    return model
+
+
+# ============================================
+# VIDEO PROCESSING
+# ============================================
+def process_video(video_path, num_frames=16, frame_size=(112, 112)):
+    """Extract and preprocess frames from video"""
+    cap = cv2.VideoCapture(video_path)
+    frames = []
+    
+    while True:
+        ret, frame = cap.read()
+        if not ret:
+            break
+        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+        frame = cv2.resize(frame, frame_size)
+        frames.append(frame)
+    
+    cap.release()
+    
+    if len(frames) == 0:
+        return None
+    
+    # Sample frames uniformly
+    total_frames = len(frames)
+    if total_frames >= num_frames:
+        indices = np.linspace(0, total_frames - 1, num_frames, dtype=int)
+    else:
+        indices = list(range(total_frames)) + [total_frames - 1] * (num_frames - total_frames)
+    
+    sampled_frames = [frames[i] for i in indices]
+    
+    # Convert to tensor: (T, H, W, C) -> (C, T, H, W)
+    video_tensor = np.stack(sampled_frames, axis=0)
+    video_tensor = video_tensor.transpose(3, 0, 1, 2)
+    video_tensor = video_tensor.astype(np.float32) / 255.0
+    
+    # Normalize
+    mean = np.array([0.485, 0.456, 0.406]).reshape(3, 1, 1, 1)
+    std = np.array([0.229, 0.224, 0.225]).reshape(3, 1, 1, 1)
+    video_tensor = (video_tensor - mean) / std
+    
+    # Add batch dimension
+    video_tensor = torch.from_numpy(video_tensor).unsqueeze(0).float()
+    
+    return video_tensor
+
+
+# ============================================
+# PREDICTION
+# ============================================
+def predict(model, video_tensor):
+    """Run prediction on video"""
+    with torch.no_grad():
+        outputs = model(video_tensor)
+        probs = torch.softmax(outputs, dim=1)
+        pred_class = torch.argmax(probs, dim=1).item()
+        confidence = probs[0][pred_class].item()
+    
+    return pred_class, confidence, probs[0].numpy()
+
+
+# ============================================
+# STREAMLIT APP
+# ============================================
+def main():
+    st.set_page_config(
+        page_title="Violence Detection",
+        page_icon="🎥",
+        layout="centered"
+    )
+    
+    st.title("🎥 Violence Detection in Videos")
+    st.markdown("**3D CNN Model trained on RWF-2000 Dataset**")
+    st.markdown("---")
+    
+    # Load model
+    with st.spinner("Loading model..."):
+        model = load_model()
+    st.success("✅ Model loaded!")
+    
+    # File uploader
+    st.markdown("### Upload a Video")
+    uploaded_file = st.file_uploader(
+        "Choose a video file (AVI, MP4, MKV)",
+        type=['avi', 'mp4', 'mkv']
+    )
+    
+    if uploaded_file is not None:
+        # Save uploaded file temporarily
+        with tempfile.NamedTemporaryFile(delete=False, suffix='.avi') as tmp_file:
+            tmp_file.write(uploaded_file.read())
+            tmp_path = tmp_file.name
+        
+        # Display video
+        st.video(uploaded_file)
+        
+        # Process and predict
+        if st.button("🔍 Analyze Video", type="primary"):
+            with st.spinner("Processing video..."):
+                # Process video
+                video_tensor = process_video(tmp_path)
+                
+                if video_tensor is None:
+                    st.error("❌ Could not process video. Please try another file.")
+                else:
+                    # Predict
+                    pred_class, confidence, probs = predict(model, video_tensor)
+                    
+                    # Display results
+                    st.markdown("---")
+                    st.markdown("### 📊 Results")
+                    
+                    col1, col2 = st.columns(2)
+                    
+                    with col1:
+                        if pred_class == 1:
+                            st.error("⚠️ **VIOLENCE DETECTED**")
+                        else:
+                            st.success("✅ **NO VIOLENCE**")
+                    
+                    with col2:
+                        st.metric("Confidence", f"{confidence * 100:.1f}%")
+                    
+                    # Probability bars
+                    st.markdown("### Probability Distribution")
+                    
+                    col1, col2 = st.columns(2)
+                    with col1:
+                        st.markdown("**Non-Violence**")
+                        st.progress(float(probs[0]))
+                        st.write(f"{probs[0] * 100:.1f}%")
+                    
+                    with col2:
+                        st.markdown("**Violence**")
+                        st.progress(float(probs[1]))
+                        st.write(f"{probs[1] * 100:.1f}%")
+            
+            # Cleanup
+            os.unlink(tmp_path)
+    
+    # Footer
+    st.markdown("---")
+    st.markdown(
+        """
+        <div style='text-align: center; color: gray;'>
+        Model: 3D CNN | Dataset: RWF-2000 | 
+        <a href='https://huggingface.co/santa47/violence-detection-3dcnn'>Model Card</a>
+        </div>
+        """,
+        unsafe_allow_html=True
+    )
+
+
+if __name__ == "__main__":
+    main()

requirements (2).txt

@@ -0,0 +1,5 @@
+streamlit==1.28.0
+torch==2.0.1
+numpy==1.24.3
+opencv-python-headless==4.8.0.76
+huggingface_hub==0.17.3