updat

detectron2/data/datasets/builtin.py

@@ -32,6 +32,12 @@ from .pascal_voc import register_pascal_voc, register_pascal_voc_wsl, register_p
 # ==== Predefined datasets and splits for COCO ==========
 
 _PREDEFINED_SPLITS_COCO = {}
+_PREDEFINED_SPLITS_COCO["voc"] = {
+    "voc_2012_seg_train": ("VOC2012/JPEGImages", "VOC2012/voc_2012_train_instance.json"),
+    "voc_2012_seg_val": ("VOC2012/JPEGImages", "VOC2012/voc_2012_val_instance.json"),
+    "voc_2012_seg_train_irn": ("VOC2012/JPEGImages", "VOC2012/voc_2012_train_instance_irn.json"),
+    "sbd_seg": ("VOC_SBD/images", "VOC2012/sbd_9118_instance.json")
+}
 _PREDEFINED_SPLITS_COCO["coco"] = {
     "coco_2014_train": ("coco/train2014", "coco/annotations/instances_train2014.json"),
     "coco_2014_val": ("coco/val2014", "coco/annotations/instances_val2014.json"),
@@ -52,10 +58,6 @@ _PREDEFINED_SPLITS_COCO["coco"] = {
     "coco_2017_gt_random_30_1": ("coco2017/train2017", "coco2017/annotations/random30_1.json"),
     "coco_2017_gt_random_30_2": ("coco2017/train2017", "coco2017/annotations/random30_2.json"),
     "coco_2017_gt_random_30_3": ("coco2017/train2017", "coco2017/annotations/random30_3.json"),
-    "voc_2012_seg_train": ("VOC2012/JPEGImages", "VOC2012/voc_2012_train_instance.json"),
-    "voc_2012_seg_val": ("VOC2012/JPEGImages", "VOC2012/voc_2012_val_instance.json"),
-    "voc_2012_seg_train_irn": ("VOC2012/JPEGImages", "VOC2012/voc_2012_train_instance_irn.json"),
-    "sbd_seg": ("VOC_SBD/images", "VOC2012/sbd_9118_instance.json")
 }
 
 _PREDEFINED_SPLITS_COCO["coco_person"] = {

detectron2/data/datasets/builtin_meta.py

@@ -13,7 +13,7 @@ The only goal is to allow users who don't have these dataset to use pre-trained
 Users don't have to download a COCO json (which contains metadata), in order to visualize a
 COCO model (with correct class names and colors).
 """
-
+import numpy as np
 
 # All coco categories, together with their nice-looking visualization colors
 # It's from https://github.com/cocodataset/panopticapi/blob/master/panoptic_coco_categories.json
@@ -231,6 +231,70 @@ ADE20K_SEM_SEG_CATEGORIES = [
 # After processed by `prepare_ade20k_sem_seg.py`, id 255 means ignore
 # fmt: on
 
+VOC_C = 21
+
+def uint82bin(n, count=8):
+    """returns the binary of integer n, count refers to amount of bits""" ""
+    return "".join([str((n >> y) & 1) for y in range(count - 1, -1, -1)])
+
+def labelcolormap(N):
+    cmap = np.zeros((N, 3), dtype=np.uint8)
+    for i in range(N):
+        r = 0
+        g = 0
+        b = 0
+        id = i
+        for j in range(7):
+            str_id = uint82bin(id)
+            r = r ^ (np.uint8(str_id[-1]) << (7 - j))
+            g = g ^ (np.uint8(str_id[-2]) << (7 - j))
+            b = b ^ (np.uint8(str_id[-3]) << (7 - j))
+            id = id >> 3
+        cmap[i, 0] = r
+        cmap[i, 1] = g
+        cmap[i, 2] = b
+    return cmap
+
+
+voc_cmaps = labelcolormap(VOC_C).tolist()
+
+VOC_CATEGORIES = [
+    {"id": 1, "name": "aeroplane", "isthing": 1, "color": voc_cmaps[1]},
+    {"id": 2, "name": "bicycle", "isthing": 1, "color": voc_cmaps[2]},
+    {"id": 3, "name": "bird", "isthing": 1, "color": voc_cmaps[3]},
+    {"id": 4, "name": "boat", "isthing": 1, "color": voc_cmaps[4]},
+    {"id": 5, "name": "bottle", "isthing": 1, "color": voc_cmaps[5]},
+    {"id": 6, "name": "bus", "isthing": 1, "color": voc_cmaps[6]},
+    {"id": 7, "name": "car", "isthing": 1, "color": voc_cmaps[7]},
+    {"id": 8, "name": "cat", "isthing": 1, "color": voc_cmaps[8]},
+    {"id": 9, "name": "chair", "isthing": 1, "color": voc_cmaps[9]},
+    {"id": 10, "name": "cow", "isthing": 1, "color": voc_cmaps[10]},
+    {"id": 11, "name": "diningtable", "isthing": 1, "color": voc_cmaps[11]},
+    {"id": 12, "name": "dog", "isthing": 1, "color": voc_cmaps[12]},
+    {"id": 13, "name": "horse", "isthing": 1, "color": voc_cmaps[13]},
+    {"id": 14, "name": "motorbike", "isthing": 1, "color": voc_cmaps[14]},
+    {"id": 15, "name": "person", "isthing": 1, "color": voc_cmaps[15]},
+    {"id": 16, "name": "pottedplant", "isthing": 1, "color": voc_cmaps[16]},
+    {"id": 17, "name": "sheep", "isthing": 1, "color": voc_cmaps[17]},
+    {"id": 18, "name": "sofa", "isthing": 1, "color": voc_cmaps[18]},
+    {"id": 19, "name": "train", "isthing": 1, "color": voc_cmaps[19]},
+    {"id": 20, "name": "tvmonitor", "isthing": 1, "color": voc_cmaps[20]},
+    {"id": 21, "name": "background", "isthing": 0, "color": [255, 255, 255]},
+]
+
+def _get_voc_instances_meta():
+    thing_ids = [k["id"] for k in VOC_CATEGORIES if k["isthing"] == 1]
+    thing_colors = [k["color"] for k in VOC_CATEGORIES if k["isthing"] == 1]
+    assert len(thing_ids) == 20, len(thing_ids)
+    # Mapping from the incontiguous VOC category id to an id in [0, 79]
+    thing_dataset_id_to_contiguous_id = {k: i for i, k in enumerate(thing_ids)}
+    thing_classes = [k["name"] for k in VOC_CATEGORIES if k["isthing"] == 1]
+    ret = {
+        "thing_dataset_id_to_contiguous_id": thing_dataset_id_to_contiguous_id,
+        "thing_classes": thing_classes,
+        "thing_colors": thing_colors,
+    }
+    return ret
 
 def _get_coco_instances_meta():
     thing_ids = [k["id"] for k in COCO_CATEGORIES if k["isthing"] == 1]
@@ -281,6 +345,8 @@ def _get_coco_panoptic_separated_meta():
 
 
 def _get_builtin_metadata(dataset_name):
+    if dataset_name == "voc":
+        return _get_voc_instances_meta()
     if dataset_name == "coco":
         return _get_coco_instances_meta()
     if dataset_name == "coco_wsl":