Primary competition visual

GeoAI Challenge for Cropland Mapping by ITU

Helping Iran (Islamic Republic of), Sudan
and 1 other country
  • Iran (Islamic Republic of)
  • Sudan
  • Afghanistan
  • Scroll to see more
$4 000 USD
Challenge completed ~2 years ago
Classification
362 joined
74 active
Starti
Jun 30, 23
Closei
Oct 05, 23
Reveali
Oct 05, 23
Zindi Please Respond - Understanding the Competition
Help · 11 Jul 2023, 00:03 · 2

@Zindi Having gone through the dataset and plotted the coordinates on Folium. I observed a few things which are unclear to me. I was hoping Zindi could help ease things by explaining in detail what they mean in each of the sections listed below;

1. In the About Section, it is stated "For the test region in Afghanistan, the participants will have to collect the imagery independently".

Question: Does this mean we are to create a test dataset for the Afghanistan region, which we would then make predictions on using our model?

2. In the About Section - it is stated "For the cropland mapping in Afghanistan, only training samples collected in 2022 can be used to train the classification model, and the model is then used to identify cropland extent in 2023"

Question: Upon observing the data, I discovered the coordinates in the test dataset are only mapped to Sudan and Iran (excluding Afghanistan). So are we training our model with training samples from Afghanistan, Sudan, and Iran for the year 2022, and testing our model on test samples from Sudan and Iran only for the year 2023?

3. In the Evaluation Section - it is stated "An evaluation jury will be set up to score each participant considering the three aspects listed below:

1. Accuracy assessment of the cropland extent maps using test samples, and the balance between training sample dataset and classification accuracies."

Question: What do you mean by the "balance between training sample dataset"?

4. In the Evaluation Section - it is stated "Checklist for data submission

1. The cropland extent of the test regions at 10m spatial resolution in the three test regions, with the following further specifications:

a) For Afghanistan (Nangarhar province): temporal cropland extent distribution of May in both 2022 and 2023.

b) For Iran and Sudan (both with one-degree by one-degree region): cropland extent distribution during the time period July 2019 ~ June 2020."

Question: Could you further explain what you mean in this section? For example, the test dataset does not contain locations in Afghanistan, but you requested the cropland extent distribution be made available for Afghanistan.

Thank you, as I look forward to your positive response :)

Discussion 2 answers

Question 3: the balance between "balance between training sample dataset and classification accuracies" indices if acquiring similar classification accuracies, the less training samples means better performance.

27 Jul 2023, 13:24
Upvotes 0

Question 4: I think the test data contain AFG now.

27 Jul 2023, 13:35
Upvotes 0