import pandas as pd

def load_user_metadata():
    # Load user metadata from CSV file
    user_metadata_path = "public_dataset/user_metadata_raw.csv"
    user_metadata = pd.read_csv(user_metadata_path)
    return user_metadata

# remove rows
remove_row_dict = {
    "user_id": ["ensemble_all","ed52331460c0479cb52d92bb6feb4bf7"],
}

remove_columns = [
    "is_ensemble_user", "user_group", "event_count", "images_annotated_real", "annotated_pixels_count_union", "gt_pixels_count_union", "annotated_pixels_count_sum_objects", "gt_pixels_count_sum_objects", "avg_polygon_confidence", "greedy_bbox_match_precision", "greedy_bbox_match_recall", "greedy_bbox_match_f1"
]
remove_columns += "greedy_bbox_match_mean_iou,coco_bbox_detection_ap,coco_bbox_detection_ap50,coco_bbox_detection_ap75,coco_bbox_detection_ar,coco_instance_segmentation_ap,coco_instance_segmentation_ap50,coco_instance_segmentation_ap75,coco_instance_segmentation_ar,semantic_segmentation_precision,semantic_segmentation_recall,semantic_segmentation_f1,semantic_segmentation_miou".split(",")

# remove rows from user_metadata DataFrame
user_metadata = load_user_metadata()
for column, values in remove_row_dict.items():
    user_metadata = user_metadata[~user_metadata[column].isin(values)]
    
user_metadata = user_metadata.drop(columns=remove_columns)

# save the cleaned user metadata to a new CSV file
cleaned_user_metadata_path = "public_dataset/user_metadata.csv"
user_metadata.to_csv(cleaned_user_metadata_path, index=False)

# check number of unique users in the cleaned user metadata
num_unique_users = user_metadata["user_id"].nunique()
print(f"Number of unique users in cleaned user metadata: {num_unique_users}")

# check number of unique images in the cleaned user metadata
num_unique_images = user_metadata["image_name"].nunique()
print(f"Number of unique images in cleaned user metadata: {num_unique_images}")