Adding doctest for image-segmentation pipeline.

src/transformers/pipelines/image_segmentation.py

@@ -32,6 +32,30 @@ class ImageSegmentationPipeline(Pipeline):
     Image segmentation pipeline using any `AutoModelForXXXSegmentation`. This pipeline predicts masks of objects and
     their classes.
 
+    Example:
+
+    ```python
+    >>> from transformers import pipeline
+
+    >>> segmenter = pipeline(model="facebook/detr-resnet-50-panoptic")
+    >>> segments = segmenter("https://huggingface.co/datasets/Narsil/image_dummy/raw/main/parrots.png")
+    >>> len(segments)
+    2
+
+    >>> segments[0]["label"]
+    'bird'
+
+    >>> segments[1]["label"]
+    'bird'
+
+    >>> type(segments[0]["mask"])  # This is a black and white mask showing where is the bird on the original image.
+    <class 'PIL.Image.Image'>
+
+    >>> segments[0]["mask"].size
+    (768, 512)
+    ```
+
+
     This image segmentation pipeline can currently be loaded from [`pipeline`] using the following task identifier:
     `"image-segmentation"`.