Detectron: Could not find instances2dict_with_polygons

[jianminsun]

import cityscapesscripts.evaluation.instances2dict_with_polygons as cs
I can't not find "instances2dict_with_polygons" in this repo or cityscapesscripts repo. Does anyone know where to find it? Thanks.

[ir413]

Hi @jianminsun, thanks for reporting the missing file. We will release it with the Cityscapes models and training instructions (please see issue #27 which tracks this).

In the meantime, you can obtain the instnaces2dict_with_polygons.py script by applying the following changes to the instances2dict.py script from the Cityscapes repo:

$ diff instances2dict.py instances2dict_with_polygons.py
--- instances2dict.py   2018-02-06 05:19:33.009812039 -0800
+++ instances2dict_with_polygons.py     2018-02-06 05:22:15.953031446 -0800
@@ -11,7 +11,9 @@
 sys.path.append( os.path.normpath( os.path.join( os.path.dirname( __file__ ) , '..' , 'helpers' ) ) )
 from csHelpers import *

-def instances2dict(imageFileList, verbose=False):
+import cv2
+
+def instances2dict_with_polygons(imageFileList, verbose=False):
     imgCount     = 0
     instanceDict = {}

@@ -35,9 +37,22 @@

         # Loop through all instance ids in instance image
         for instanceId in np.unique(imgNp):
+            if instanceId < 1000:
+                continue
+
             instanceObj = Instance(imgNp, instanceId)
+            instanceObj_dict = instanceObj.toDict()

-            instances[id2label[instanceObj.labelID].name].append(instanceObj.toDict())
+            if id2label[instanceObj.labelID].hasInstances:
+                mask = (imgNp == instanceId).astype(np.uint8)
+                contour, hier = cv2.findContours(
+                    mask.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
+
+                polygons = [c.reshape(-1).tolist() for c in contour]
+                instanceObj_dict['contours'] = polygons
+
+            instances[id2label[instanceObj.labelID].name].append(
+                instanceObj_dict)

         imgKey = os.path.abspath(imageFileName)
         instanceDict[imgKey] = instances

[lssily]

import cityscapesscripts.evaluation.instances2dict_with_polygons as cs
I can't not find "instances2dict_with_polygons" in this repo or cityscapesscripts repo. Does anyone know where to find it? Thanks.

Hi! Recently, I have trained a model on Cityscapes dataset. However, the perforamce is half the result reported by paper. Did that happen to you?

[zhixy]

@ir413 @lssily can you just share the generated json files for cityscapes?

[xllau]

import cityscapesscripts.evaluation.instances2dict_with_polygons as cs
I can't not find "instances2dict_with_polygons" in this repo or cityscapesscripts repo. Does anyone know where to find it? Thanks.

Hi! Recently, I have trained a model on Cityscapes dataset. However, the perforamce is half the result reported by paper. Did that happen to you?

Did you solve the problem, I also meet the problem of the val accuracy problem. It is too low to match the results in paper.

[viet2411]

Sorry I don't really understand this part of convert_cityscapes_to_coco.py:

`
def convert_cityscapes_instance_only(
data_dir, out_dir):
"""Convert from cityscapes format to COCO instance seg format - polygons"""
sets = [
'gtFine_val',
# 'gtFine_train',
# 'gtFine_test',

    # 'gtCoarse_train',
    # 'gtCoarse_val',
    # 'gtCoarse_train_extra'
]
ann_dirs = [
    'gtFine_trainvaltest/gtFine/val',
    # 'gtFine_trainvaltest/gtFine/train',
    # 'gtFine_trainvaltest/gtFine/test',

    # 'gtCoarse/train',
    # 'gtCoarse/train_extra',
    # 'gtCoarse/val'
]

`
I don't understand how I should organize the directory for dataset. What is the correct way to perform this code?
Can anyone share the generated json files for cityscapes?

[jinfagang]

Why not just given that file..... what does these diff can applying for