Discord

Checklist INAIO 2025 Stage 2 · Task 6

Drone Patrol: From Vivid Landscapes to Monochrome Missions

Classify drone images as coast or desert using labelled colour images and unlabelled grayscale images, for colour or grayscale inputs.

  • Vision
  • Image classification with domain shift (semi-supervised)

The task

Daytime colour images are labelled by experts, while night-time grayscale images are unlabelled; at patrol time images may be either. Q1 builds learn_location_type_classifier and classify_image; Q2 evaluates on public colour and grayscale test splits; Q3 is a secret-set evaluation (14 points in total).

Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.

At a glance

You get
color_images_train.csv, grayscale_images_train.csv, color_test_public.csv, grayscale_test_public.csv, processed_images.zip.
You submit
Completed notebook functions and reported metrics.
Scoring
ROC curve and AUC on each public split.
Rules
  • Make a copy of the notebook, share it with the organisers and report LLM chats and external resources used in the main test form.
Format
INAIO 2025 Stage 2, 2 March 2025, 10:00 AM – 1:00 PM IST (from the test form). Theory problems in a Google Form plus ML coding problems in Colab notebooks shared with the organisers.

Details

Year
2025, Online (proctored)
Round
Stage 2 · Task 6
Language
English
License
Not stated by the source