You will be provided with:

- **`ref_img.csv`**: Contains one reference image ID per celebrity (15 rows total)
- **`images/`**: Directory containing all 108 images as `.jpg` files, named with 3-digit IDs (e.g., `000.jpg`, `001.jpg`, etc.)

For each reference image in `ref_img.csv`, you must find all images from the dataset that **show the same celebrity** as the reference image. The reference image itself should **not** be included in your results.