Update cocostuff processing

Former-commit-id: 28bae6d46e54331a42544fe5239ad3cee2653117

README.md

@@ -109,9 +109,9 @@ pip install -r requirements.txt
 ### Training Data Preparation
 The training data consists of 4 types of data:
 
-1. Semantic segmentation datasets: [ADE20K](http://data.csail.mit.edu/places/ADEchallenge/ADEChallengeData2016.zip), COCO-Stuff [\[images\]](http://images.cocodataset.org/zips/train2017.zip) [\[labels\]](http://calvin.inf.ed.ac.uk/wp-content/uploads/data/cocostuffdataset/stuffthingmaps_trainval2017.zip), [Mapillary](https://www.mapillary.com/dataset/vistas), [PACO-LVIS](https://github.com/facebookresearch/paco/tree/main#dataset-setup), [PASCAL-Part](https://github.com/facebookresearch/VLPart/tree/main/datasets#pascal-part)
+1. Semantic segmentation datasets: [ADE20K](http://data.csail.mit.edu/places/ADEchallenge/ADEChallengeData2016.zip), [COCO-Stuff](http://calvin.inf.ed.ac.uk/wp-content/uploads/data/cocostuffdataset/stuffthingmaps_trainval2017.zip), [Mapillary](https://www.mapillary.com/dataset/vistas), [PACO-LVIS](https://github.com/facebookresearch/paco/tree/main#dataset-setup), [PASCAL-Part](https://github.com/facebookresearch/VLPart/tree/main/datasets#pascal-part), [COCO Images](http://images.cocodataset.org/zips/train2017.zip)
 
-    Note: For COCO-Stuff, we use the annotation file stuffthingmaps_trainval2017.zip. We only use the PACO-LVIS part in PACO.
+    Note: For COCO-Stuff, we use the annotation file stuffthingmaps_trainval2017.zip. We only use the PACO-LVIS part in PACO. COCO Images should be put into the `coco` directory.
 
 3. Referring segmentation datasets: [refCOCO](https://web.archive.org/web/20220413011718/https://bvisionweb1.cs.unc.edu/licheng/referit/data/refcoco.zip), [refCOCO+](https://web.archive.org/web/20220413011656/https://bvisionweb1.cs.unc.edu/licheng/referit/data/refcoco+.zip), [refCOCOg](https://web.archive.org/web/20220413012904/https://bvisionweb1.cs.unc.edu/licheng/referit/data/refcocog.zip), [refCLEF](https://web.archive.org/web/20220413011817/https://bvisionweb1.cs.unc.edu/licheng/referit/data/refclef.zip) ([saiapr_tc-12](https://web.archive.org/web/20220515000000/http://bvisionweb1.cs.unc.edu/licheng/referit/data/images/saiapr_tc-12.zip)) 
 
@@ -130,9 +130,10 @@ Download them from the above links, and organize them as follows.
 │   │   └── images
 │   ├── coco
 │   │   └── train2017
+│   │       ├── 000000000009.jpg
+│   │       └── ...
 │   ├── cocostuff
 │   │   └── train2017
-│   │       ├── 000000000009.jpg
 │   │       ├── 000000000009.png
 │   │       └── ...
 │   ├── llava_dataset

utils/sem_seg_dataset.py

@@ -80,15 +80,15 @@ def init_cocostuff(base_image_dir):
             cocostuff_classes.append(line.strip().split(": ")[-1])
     cocostuff_classes = np.array(cocostuff_classes)
     cocostuff_images = []
-    cocostuff_image_dir = glob.glob(
-        os.path.join(base_image_dir, "cocostuff", "train2017", "*.jpg")
+
+    cocostuff_labels = glob.glob(
+        os.path.join(base_image_dir, "cocostuff", "train2017", "*.png")
     )
-    for image_id in cocostuff_image_dir:
-        cocostuff_images.append(image_id)
-    cocostuff_labels = [
-        x.replace(".jpg", ".png").replace("images", "annotations")
-        for x in cocostuff_images
+    cocostuff_images = [
+        x.replace(".png", ".jpg").replace("cocostuff", "coco")
+        for x in cocostuff_labels
     ]
+
     print("cocostuff: ", len(cocostuff_images))
     return cocostuff_classes, cocostuff_images, cocostuff_labels