Skip to main content

Dataset Annotation on Jetson

To train custom vision models or improve recognition accuracy, dataset collection and annotation are critical steps. This guide explains how to collect image data and annotate it using Label Studio on NVIDIA Jetson.

1. Dataset Collection​

A common approach is to record a video and extract frames at regular intervals.

Extract Images from Video​

The following example extracts one image every 15 frames:

import cv2
import os

video_file = 'orange.mkv'
output_folder = 'orange_images'

os.makedirs(output_folder, exist_ok=True)

cap = cv2.VideoCapture(video_file)
frame_interval = 15
frame_number = 0
saved_frame = 1

while True:
    ret, frame = cap.read()
    if not ret:
        break

    if frame_number % frame_interval == 0:
        cv2.imwrite(f"{output_folder}/{saved_frame}.png", frame)
        saved_frame += 1

    frame_number += 1

cap.release()
print("Frame extraction completed.")

2. Install Label Studio​

Label Studio is an open-source data annotation platform.

Install via Docker​

docker pull heartexlabs/label-studio:latest

3. Launch Label Studio​

Set Dataset Permissions​

sudo chmod 777 ~/ultralytics/ultralytics/data

Run Label Studio Container​

sudo docker run -it   -p 8080:8080   -v ~/ultralytics/ultralytics/data:/label-studio/data   heartexlabs/label-studio:latest   label-studio

Access Label Studio​

4. Annotation Workflow​

Create Project​

  • Select Image Annotation
  • Name the project clearly

Import Images​

Upload images from your dataset directory.

5. Label Configuration Example​

<View>
  <Image name="image" value="$image"/>
  <RectangleLabels name="label" toName="image">
    <Label value="object"/>
  </RectangleLabels>
</View>

6. Export Dataset​

Recommended formats: - YOLO - COCO - Pascal VOC

YOLO Dataset Structure​

dataset/
├── images/
│   ├── train/
│   └── val/
├── labels/
│   ├── train/
│   └── val/
└── data.yaml

Example data.yaml​

path: dataset
train: images/train
val: images/val

names:
  0: object

Summary​

  • Extract images from real scenes
  • Annotate with Label Studio
  • Export to YOLO format for training
  • Use consistent naming and structure

Maintained by HemiHex for Jetson-based advanced vision workflows.