Could you give the .json file of cityscapes dataset in coco format?

[mrsempress]

Actual results

I download gtFine_trainvaltest.zip, leftImg8bit_trainvaltest.zip, leftImg8bit_trainvaltest_foggy.zip, leftImg8bit_trainval_rain.zip from https://www.cityscapes-dataset.com/. And I use tools/convert_cityscapes_to_coco.py, also succeed to convert the datasets. Then I want to use the faster rcnn in mmdetection to do the control experiment, but the results are too low. In train, I use images in leftImg8bit_trainvaltest.zip, and in val or test, I use images in leftImg8bit_trainvaltest_foggy.zip or leftImg8bit_trainval_rain.zip. So, I want to ask for the .json, or maybe I forget to do something? The results I do is :

dataset	Person	Rider	Car	Bus	Truck	Bicycle	Mcycle	Train	mAP
rainy	1.2	1.5	1.8	1.8	0.3	1.0	0	0	1.8
foggy	9.1	9.4	14.2	9.4	4.5	7.5	4.2	1.2	14.6

Expected results

I read the paper, and their results (in foggy) are:

Paper name	Person	Rider	Car	Bus	Truck	Bicycle	Mcycle	Train	mAP
A Robust Learning Approach to Domain Adaptive Object Detection	31.69	39.41	45.81	23.86	39.34	20.64	22.26	32.36	31.92
Domain Adaptive Faster R-CNN for Object Detection in the Wild	17.8	23.6	27.1	23.8	11.9	22.8	14.4	9.1	18.8

[Ronales]

@mrsempress Dear,Can you share download link for leftImg8bit_trainvaltest_foggy.zip？I have send email to official email,but I have receive no reply.Thanks in advance!

[mrsempress]

@mrsempress Dear,Can you share download link for leftImg8bit_trainvaltest_foggy.zip？I have send email to official email,but I have receive no reply.Thanks in advance!

Sorry, the only way to get leftImg8bit_trainvaltest_foggy.zip is waiting for official email. Because after he reply your email, then you can find your download page(the address also same as before) appears the dataset your ask for. Hope you get reply soon.

[Ronales]

@mrsempress Thanks！