Upload 10 files

.gitattributes

@@ -33,3 +33,4 @@ 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
+0TqQje61Hoo.3.1.mp4 filter=lfs diff=lfs merge=lfs -text

0TqQje61Hoo.3.1.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e43dc88426637b14fd369705158d3eafbd186137cdd47a7df3c5ee37302d0ba0
+size 2052375

README.md

@@ -1,19 +1,89 @@
----
-title: Camera Detector
-emoji: 🚀
-colorFrom: red
-colorTo: red
-sdk: docker
-app_port: 8501
-tags:
-- streamlit
-pinned: false
-short_description: Streamlit template space
----
-
-# Welcome to Streamlit!
-
-Edit `/src/streamlit_app.py` to customize this app to your heart's desire. :heart:
-
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
+# 📷 Camera Movement Detector
+
+This project was developed as part of the **ATP Core Talent AI Coder Challenge 2025**.  
+It is a Streamlit-based web application that analyzes uploaded videos and detects **significant camera movements** such as panning, tilting, or shaking.  
+It does **not** focus on object motion — instead, it detects changes in the global scene indicating that the camera itself moved.
+
+---
+
+## 📷 Screenshot
+
+![Camera Detector Screenshot](https://github.com/user-attachments/assets/90a2831b-2218-42a6-8172-f3ee63661157)
+
+## 🚀 Live App
+
+🔗 [Click here to try the app online](https://your-deployment-url.com)  
+(Replace this with your Streamlit or Hugging Face link after deployment)
+
+---
+
+## 🎯 Features
+
+- 📼 Upload a video (MP4, AVI, MOV)
+- 🎞️ Extracts and analyzes frames
+- 🎯 Detects significant camera movement (not object motion)
+- 💡 Clean, modern UI with ATP branding
+
+---
+
+## 🧠 Approach
+
+- **Frame Extraction:** Extracts all frames from the video using OpenCV.
+- **Feature Detection & Matching:** ORB feature detector is used to compare consecutive frames.
+- **Homography Transformation:** Calculates transformation matrix to estimate global scene shift.
+- **Movement Decision:** If the transformation deviates above a threshold, it is classified as camera movement.
+- **Bonus Logic:** Tries to filter out object motion by comparing matched keypoints and transformation consistency.
+
+---
+
+## 🛠️ Tech Stack
+
+- **Python 3.13**
+- **OpenCV**
+- **Streamlit** (for UI)
+- **Datasets** from Hugging Face (for CameraBench demo data)
+- **NumPy** for numerical processing
+
+---
+
+## 📦 Installation
+
+To run locally:
+
+```bash
+git clone https://github.com/yourusername/camera-detector.git
+cd camera-detector
+pip install -r requirements.txt
+streamlit run app.py
+```
+
+## 📁 File Structure
+
+```bash
+camera-detector/
+├── app.py                 # Streamlit web UI
+├── movement_detector.py   # Core detection logic
+├── requirements.txt       # Dependencies
+├── README.md              # This file
+```
+
+### 📸 Example Output
+
+Input: Sample video with camera shake
+
+#### Output:
+
+```bash
+
+📌 Detected camera movement at frame indices: [6, 7, 8, 10, 15, 22, 23, 24]
+```
+## 💬 Acknowledgements
+
+* Hugging Face CameraBench Dataset
+* ATP Core Talent Team for organizing the challenge
+
+## 🧑‍💻 Developed by
+
+* Saadet Elizaveta Babal
+* Candidate – ATP Core Talent AI Coder Challenge 2025
+

README2.md

@@ -0,0 +1,95 @@
+# ATP Core Talent 2025
+# Core Talent AI Coder Challenge: Camera Movement Detection
+
+**Detecting Significant Camera Movement Using Image Recognition**
+
+---
+
+## Scenario
+
+Imagine you are tasked with building a component for a smart camera system. Your goal is to detect **significant movement**—for example, if someone moves or tilts the camera or if the entire camera is knocked or shifted. This is different from simply detecting moving objects in the scene.
+
+---
+
+## Requirements
+
+1. **Input:**
+
+   * A sequence of images or frames (at least 10-20), simulating a fixed camera, with some frames representing significant camera movement (tilt, pan, large translation), and others showing a static scene or minor background/object motion.
+   * You may use public datasets, generate synthetic data, or simulate with your own webcam.
+
+     * Example: [CameraBench Dataset on Hugging Face](https://huggingface.co/datasets/syCen/CameraBench)
+2. **Task:**
+
+   * Build an algorithm (**Python preferred**) that analyzes consecutive frames and detects when significant camera movement occurs.
+   * Output a list of frames (by index/number) where significant movement is detected.
+3. **Expected Features:**
+
+   * **Basic:** Frame differencing or feature matching to detect large global shifts (e.g., using OpenCV’s ORB/SIFT/SURF, optical flow, or homography).
+   * **Bonus:** Distinguish between camera movement and object movement within the scene (e.g., use keypoint matching, estimate transformation matrices, etc.).
+4. **Deployment:**
+
+   * Wrap your solution in a small web app (**Streamlit, Gradio, or Flask**) that allows the user to upload a sequence of images (or a video), runs the detection, and displays the result.
+   * Deploy the app on a public platform (**Vercel, Streamlit Cloud, Hugging Face Spaces**, etc.)
+5. **Deliverables:**
+
+   * Public app URL
+   * GitHub repo (with code and requirements.txt)
+   * README (explaining your approach, dataset, and how to use the app)
+
+     * **Sample README Outline:**
+
+       * Overview of your approach and movement detection logic
+       * Any challenges or assumptions
+       * How to run the app locally
+       * Link to the live app
+       * Example input/output screenshots
+   * AI Prompts or Chat History (if used for support)
+
+---
+
+## Evaluation Rubric
+
+| Criteria           | Points | Details                                                                                    |
+| ------------------ | ------ | ------------------------------------------------------------------------------------------ |
+| **Correctness**    | 5      | Accurately detects significant camera movement; low false positives/negatives.             |
+| **Implementation** | 5      | Clean code, good use of OpenCV or relevant libraries, modular structure.                   |
+| **Deployment**     | 5      | App is online, easy to use, and functions as described.                                    |
+| **Innovation**     | 3      | Advanced techniques (feature matching, transformation estimation, clear object vs camera). |
+| **Documentation**  | 2      | Clear README, instructions, and concise explanation of method/logic.                       |
+
+---
+
+## Suggested Stack
+
+* **Python** or **C#**
+* **OpenCV** for computer vision
+* **Streamlit**, **Gradio**, or a **shadcn-powered Vercel site** for quick web UI
+* **GitHub** for code repo, **Streamlit Cloud**, **Hugging Face Spaces**, or **Vercel** for deployment
+
+---
+
+# 📋 Candidate Instructions
+
+1. **Fork this repository** (or start your own repository with the same structure).
+2. **Implement your movement detection algorithm** in `movement_detector.py`.
+3. **Develop a simple web app** (`app.py`) that allows users to upload images/sequences and view detection results.
+4. **Deploy your app** on a public platform (e.g., Streamlit Cloud, Hugging Face Spaces, Vercel, Heroku) and **share both your deployed app URL and GitHub repository link**.
+5. **Document your work**: Include a `README.md` that explains your approach, how to run your code, and sample results (with screenshots or example outputs).
+
+---
+
+**Deadline:**
+🕓 **27.06.2025**
+
+---
+
+**Plagiarism Policy:**
+
+* This must be **individual, AI-powered work**.
+* You may use open-source libraries, but you **must cite** all external resources and code snippets.
+* Do not submit work copied from others or from the internet without proper acknowledgment.
+
+---
+
+**Good luck! Show us your best hands-on AI skills!**

app.py

@@ -0,0 +1,128 @@
+import os
+
+# Streamlit config dosyasını kullanıcının home dizininde oluştur
+config_dir = os.path.join(os.path.expanduser("~"), ".streamlit")
+os.makedirs(config_dir, exist_ok=True)
+
+config_path = os.path.join(config_dir, "config.toml")
+with open(config_path, "w") as f:
+    f.write("""
+[server]
+headless = true
+port = $PORT
+enableCORS = false
+""")
+
+import streamlit as st
+import tempfile
+from movement_detector import extract_frames, detect_camera_movement
+
+st.set_page_config(page_title="Camera Movement Detector", layout="centered")
+
+# CSS + HTML (Profesyonel Tasarım)
+st.markdown("""
+    <style>
+    body {
+        font-family: 'Segoe UI', sans-serif;
+    }
+
+    .stApp {
+        background-color: #f4f4f4;
+        padding-top: 30px;
+    }
+
+    .logo-container {
+        display: flex;
+        justify-content: center;
+        align-items: center;
+        margin-bottom: 15px;
+    }
+
+    .logo-container img {
+        width: 60px;
+        height: auto;
+        opacity: 0.9;
+    }
+
+    .card {
+        background-color: white;
+        border-radius: 12px;
+        padding: 30px;
+        max-width: 700px;
+        margin: auto;
+        box-shadow: 0 8px 20px rgba(0,0,0,0.08);
+    }
+
+    .title {
+        text-align: center;
+        font-size: 26px;
+        color: #333;
+        margin-bottom: 10px;
+        font-weight: 600;
+    }
+
+    .description {
+        text-align: center;
+        font-size: 16px;
+        color: #666;
+        margin-bottom: 25px;
+    }
+
+    .stButton>button {
+        background-color: #ff4b4b;
+        color: white;
+        border-radius: 6px;
+        padding: 8px 20px;
+        font-weight: bold;
+    }
+
+    .stFileUploader label {
+        font-weight: 500;
+        color: #444;
+    }
+
+    /* 🔧 Info ve spinner kutularının metin rengini siyah yap */
+    .stAlert > div {
+        color: black !important;
+    }
+
+    .stSpinner > div {
+        color: black !important;
+    }
+
+    </style>
+""", unsafe_allow_html=True)
+
+# 🔹 Logo
+st.markdown("""
+<div class="logo-container">
+    <img src="https://i.hizliresim.com/khnekdt.png" alt="ATP Logo">
+</div>
+""", unsafe_allow_html=True)
+
+# 🔹 İçerik Kartı
+st.markdown('<div class="card">', unsafe_allow_html=True)
+
+st.markdown('<div class="title">📷 Camera Movement Detector</div>', unsafe_allow_html=True)
+st.markdown('<div class="description">Upload a video to detect <b>significant camera movement</b> (not object motion).</div>', unsafe_allow_html=True)
+
+# 🔸 Dosya yükleyici
+uploaded_video = st.file_uploader("🎞️ Upload your video", type=["mp4", "avi", "mov"])
+
+# 🔸 Analiz işlemi
+if uploaded_video:
+    with tempfile.NamedTemporaryFile(delete=False) as temp_file:
+        temp_file.write(uploaded_video.read())
+        temp_video_path = temp_file.name
+
+    st.info("Extracting frames...")
+    frames = extract_frames(temp_video_path)
+
+    st.info("Analyzing for camera movement...")
+    with st.spinner("Running detection..."):
+        indices = detect_camera_movement(frames)
+
+    st.success("✅ Detection complete.")
+    st.markdown(f"<div class='description'>📌 Movement detected at frames:<br><code>{indices}</code></div>", unsafe_allow_html=True)
+
+st.markdown('</div>', unsafe_allow_html=True)

download_video.py

@@ -0,0 +1,18 @@
+import requests
+
+# Dataset path (elde ettiğin path)
+video_path = "videos/0TqQje61Hoo.3.1.mp4"
+
+# Hugging Face URL
+base_url = "https://huggingface.co/datasets/syCen/CameraBench/resolve/main/"
+video_url = base_url + video_path
+
+# Kaydedilecek dosya adı
+output_filename = video_path.split("/")[-1]
+
+# İndirme işlemi
+response = requests.get(video_url)
+with open(output_filename, "wb") as f:
+    f.write(response.content)
+
+print(f"✅ Video indirildi: {output_filename}")

get_videos.py

@@ -0,0 +1,12 @@
+from datasets import load_dataset
+
+# Dataset'i yükle
+ds = load_dataset("syCen/CameraBench")
+test_data = ds["test"]
+
+# İlk 3 video bilgisini yazdır
+for i in range(3):
+    print(f"Caption: {test_data[i]['caption']}")
+    print(f"Label: {test_data[i]['labels']}")
+    print(f"Path: {test_data[i]['path']}")
+    print()

movement_detector.py

@@ -0,0 +1,49 @@
+import cv2
+import numpy as np
+
+def extract_frames(video_path, resize=(640, 480)):
+    cap = cv2.VideoCapture(video_path)
+    frames = []
+    while cap.isOpened():
+        ret, frame = cap.read()
+        if not ret:
+            break
+        if resize:
+            frame = cv2.resize(frame, resize)
+        frames.append(cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY))
+    cap.release()
+    return frames
+
+def detect_camera_movement(frames, threshold=15):
+    orb = cv2.ORB_create()
+    movement_indices = []
+
+    for i in range(len(frames) - 1):
+        kp1, des1 = orb.detectAndCompute(frames[i], None)
+        kp2, des2 = orb.detectAndCompute(frames[i+1], None)
+
+        if des1 is None or des2 is None:
+            continue
+
+        bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
+        matches = bf.match(des1, des2)
+
+        if len(matches) < 10:
+            continue
+
+        src_pts = np.float32([kp1[m.queryIdx].pt for m in matches]).reshape(-1, 1, 2)
+        dst_pts = np.float32([kp2[m.trainIdx].pt for m in matches]).reshape(-1, 1, 2)
+
+        # Homografi matrisi hesaplanıyor
+        H, mask = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)
+
+        if H is not None:
+            # Homografi’den dönüşüm şiddetini tahmin ediyoruz
+            dx = H[0, 2]
+            dy = H[1, 2]
+            movement_magnitude = np.sqrt(dx**2 + dy**2)
+
+            if movement_magnitude > threshold:
+                movement_indices.append(i+1)
+
+    return movement_indices

requirements.txt

@@ -1,3 +1,5 @@
-altair
-pandas
-streamlit
+streamlit>=1.35.0
+opencv-python>=4.9.0
+numpy>=1.24.0
+Pillow>=10.0.0
+datasets>=3.6.0

test_dataset.py

@@ -0,0 +1,5 @@
+from datasets import load_dataset
+
+# Hugging Face'den CameraBench veri setini indir
+ds = load_dataset("syCen/CameraBench")
+print(ds)