Number of categories in concatenated Images
Autorius: Mihai Nan
Vidutinis
Jūsų geriausias rezultatas: Nepasiekiama
Užduoties aprašymas
🏞️ Number of Categories in a Concatenated Image
For this task, you need to implement a model capable of estimating the number of distinct categories in an image created by concatenating several smaller images.
🔹 Dataset
The dataset is provided as images with associated CSV files:
-
train.csv: contains examples for training the model
Column Type Description SampleID integer Unique identifier for each example ImagePath string Path to the concatenated image Label integer Number of distinct categories in the image -
test.csv: contains examples for prediction
Column Type Description SampleID integer Unique identifier ImagePath string Path to the concatenated image
⚙️ Evaluation Metric
The score for each example is calculated as follows:
- If the predicted value equals the true value → 1 point
- If the predicted value differs from the true value by one (+1/-1) → 0.5 points
- Otherwise → 0 points
📨 Submission File Format
The submission file must be a CSV with two columns:
| Column | Type | Description |
|---|---|---|
| SampleID | integer | Unique identifier of the row in the test set |
| PredictedLabel | integer | Predicted number of distinct categories in the image |
🔹 Example submission.csv
| SampleID | PredictedLabel |
|---|---|
| 1 | 4 |
| 2 | 2 |
| 3 | 3 |
🖼️ Example Interpretation of the Label
If a concatenated image contains multiple objects, the label represents the number of distinct categories.
🏞️ Example
- The image contains: a dog, a car, and another dog
- Number of distinct categories = 2 (category "dog" and category "car")
Thus, the model must recognize the different classes in the image and return an integer corresponding to the number of distinct classes.