null
Lookup-Jet: Multimodal Aviation Tracking Dataset
Overview
The Lookup-Jet Dataset is an industry-grade, sensor-fusion dataset designed for advanced computer vision and machine learning tasks. It combines high-resolution visual tracking (bounding boxes and polygon segmentations) of aircraft with synchronized ADS-B kinematics, environmental conditions, and astronomical data.
This dataset is pre-formatted for immediate deployment across standard ML frameworks, supporting both YOLO (edge/real-time) and COCO (research/heavyweight) architectures.
📂 Directory Structure
my_video_dataset/
├── README.md # This document
├── data_dictionary.md # Human-readable field definitions
├── data_dictionary.json # Machine-readable field schema
├── dataset.yaml # YOLO configuration file
├── telemetry.csv # Master sensor-fusion telemetry (Pandas-ready)
├── images/ # Extracted, unadulterated .jpg video frames
├── labels/ # Normalized YOLO .txt segmentation masks
├── annotations/
│ └── instances_default.json # Master COCO format segmentation dictionary
└── qa/ # Visual QA frames with burned-in annotations
🚀 Quick Start / Usage
1. Vision Tasks: YOLO (Ultralytics)
The dataset is natively structured for YOLOv8/v11 segmentation tasks. To begin training immediately:
yolo task=segment mode=train data=dataset.yaml model=yolov8n-seg.pt epochs=100
2. Vision Tasks: COCO (Detectron2 / PyTorch)
For frameworks expecting Microsoft COCO formatting, use the master JSON located in the annotations/ folder.
- Categories:
0: airplane
3. Sensor Fusion / Time-Series (Pandas)
All telemetry and metadata are flattened into telemetry.csv, synchronized perfectly to the frame_num.
import pandas as pd
# Load the entire telemetry timeline
df = pd.read_csv('telemetry.csv')
# Example: Filter for frames where the plane is banking left and altitude > 5000ft
banking_left = df[(df['roll'] < 0) & (df['alt_geom'] > 5000)]
- Downloads last month
- -