kaityc06/WildlifeReID-10k_masked

WildlifeReID-10k Masked

Segmentation masks for the WildlifeReID-10k dataset, stored as RLE-encoded masks in a single Parquet file.

File

File	Description
masks.parquet	89,002 rows — one per image — with RLE mask, score, species label, individual identity, and bbox

Schema

Column	Type	Description
dataset	str	Always WildlifeReID-10k
text_prompt	str	Text prompt used to generate the mask (may differ from species — e.g. nyala → antelope)
mask_rle_counts	str	RLE-encoded mask (COCO format)
mask_rle_height	int	Height used for RLE decoding
mask_rle_width	int	Width used for RLE decoding
mask_score	float	Mask confidence score
image_width	int	Original image width
image_height	int	Original image height
species	str	Species name (one of 22 classes, e.g. tiger, sea turtle, whaleshark)
identity	str	Individual animal identity label (9,406 unique IDs, e.g. ATRW_99)
bbox	str	Bounding box [x, y, w, h] in pixels, or null if unavailable

Loading the masks

import pandas as pd

df = pd.read_parquet("hf://datasets/kaityc06/WildlifeReID-10k_masked/masks.parquet")

Retrieving the original image from an identity

The identity column matches the identity column in the source dataset's metadata.csv, which maps each identity to a relative image path.

import os
import pandas as pd
from PIL import Image

# Path to your local WildlifeReID-10k cache (must contain metadata.csv and images/)
WILDLIFEREID_CACHE = "/path/to/wildlifereid-10k/data"

# Build identity → absolute path mapping (one row per image)
meta = pd.read_csv(os.path.join(WILDLIFEREID_CACHE, "metadata.csv"), low_memory=False)
identity_to_paths = meta.groupby("identity")["path"].apply(list).to_dict()

# Look up any identity from the parquet
identity = "ATRW_99"
img_paths = identity_to_paths[identity]
img = Image.open(os.path.join(WILDLIFEREID_CACHE, img_paths[0]))

Decoding a mask

import numpy as np
from pycocotools import mask as mask_utils

row = df.iloc[0]

rle = {
    "counts": row["mask_rle_counts"],
    "size": [row["mask_rle_height"], row["mask_rle_width"]],
}
binary_mask = mask_utils.decode(rle)  # numpy array, shape (H, W), dtype uint8