Intelligent-Systems-Project · no06

no06 converts raw videos into pose-based colour features, clusters them with t-SNE + k-means and offers a smooth, single-page UX for exploration. This is an application created in Intelligent Systems Project 1 at KDIX.

---

✨ Features

Area	What you get
Upload & storage	AJAX form, metadata in SQLite (Video), files under no06/media/.
Feature extraction	1 fps sampling → YOLOv8-Pose on shoulders & hips → 12-dim RGB vector.
Clustering	t-SNE (3-D) + k-means (k = 5) rendered via Plotly.
Interactive UI	Click a point to preview, or run a slide-show; pop-up video player in upload list.
BBox detector	Optional YOLOv8 object-detector video overlay.
Safe deletion	CSRF-protected POST removes a video and every derived asset.

---

🚀 Installation & Quick-start

1. Prerequisites

• Python 3.10 or 3.11
• ffmpeg (for BBox video) — brew install ffmpeg or apt install ffmpeg

2. Set up

# clone & create a virtual environment
git clone https://github.com/<YOU>/Intelligent-Systems-Project.git
cd Intelligent-Systems-Project
python -m venv .venv && source .venv/bin/activate

# install requirements
pip install -r requirements.txt

# download YOLO weights (≈ 90 MB total)
python - <<'PY'
from ultralytics import YOLO
YOLO('yolov8n-pose.pt'); YOLO('yolov8n.pt')
PY

# initialise the database
python manage.py migrate

# run the dev server
python manage.py runserver
Browse to http://127.0.0.1:8000/ — you’re ready to upload a video.

🔗 Key AJAX / REST Endpoints

Verb	Path	Params	Returns
POST	/upload/	name, video_file	{"message": "..."}"
POST	/deleteVideo/<vid>/	—	Redirect → / (HTML) or JSON
GET	/generateData/	video_ids=1,2	JSON status
GET	/make_distribution/	video_ids=…	JSON status
GET	/showDistribution/	video_ids=…	Plotly figure JSON
GET	/bboxDetect/	video_id=	JSON {ok}
GET	/getVideo/	path=	MP4 stream
GET	/getThumbList/	video_id=	JSON list of thumbnails
GET	/getImage/	path=	JPEG

🐞 Troubleshooting

Issue	Remedy
YOLO torch.load warning	Harmless — upstream note only.
ffmpeg missing	Install via package manager.
Upload fails > 100 MB	Raise DATA_UPLOAD_MAX_MEMORY_SIZE or front-end proxy limit.
SSL cert error on macOS	Run Install Certificates.command bundled with Python.