autonue / public-code Goto Github PK

Now, I had trained the semantic segmentation model on the IDD dataset. I want to test it using test set. My data format is:[0,25] for 26 class.
How convert the data format so that I can submit test set result.

Wanted to understand where is the annotations file for the IDD Detection data set available on .

args variable that is inaccessible in preperation/createLabels.py in process_folder function I'm working on semantic segmentation so for now I hardcoded for myself for id_type, color and instance.

I am using google colab and while running the createLabels.py I am getting an ImportError message as:

Traceback (most recent call last):
File "preperation/createLabels.py", line 9, in
from scipy.misc import imread, imsave
ImportError: cannot import name 'imread'

I tried looking into the scipy docs for version '1.4.1' and I am not able to find any imread or imsave.

I am working on evaluating my instance segmentation output on the given dataset. I don't seem to understand the format of the output required by the provided script.

Specifically the below part of the evaluate_instance_segmentation.py script:

# To run this script, make sure that your results contain text files
# (one for each test set image) with the content:
#   relPathPrediction1 labelIDPrediction1 confidencePrediction1
#   relPathPrediction2 labelIDPrediction2 confidencePrediction2
#   relPathPrediction3 labelIDPrediction3 confidencePrediction3
#   ... 

What should the LabelIDPrediction be? Is it the class ID? Or is it the encoded ID obtained similar to using ID_TYPE=id in createLabels.py? One more concern is the relPathPrediction in the above text format. It says that the path should be relative to the root directory. I need help understanding what root directory is being referred to, since I could not find the default variable.

The evaluate_instance_segmentation.py takes input defined as follows:

Do multiple files need to be created (one for each label) per image?

Can you provide such sample file for validation images in the dataset so as to check the working of the script. Or for at least one image? Or tell me how do I generate such file from *_gtFine_instanceids.png?

Hi,

I was trying to generate instance labels for the new release (part 2) of the IDD. However, I observed that polygons in the json file for certain classes do not have any coordinates (i.e. the labelled bounding boxes are empty) which keeps giving an error. I didn't face such error in the first part of the dataset.

Please suggest some workaround for this issue.

Thanks.

ModuleNotFoundError: No module named 'anue_labels'

Hi, I work on instance segmentation. However, I find it hard to make evaluation by your code. It works well on coco-liking evaluation.

Firstly, how could I get gtFine_instanceids.png in 658 line on evaluate_instance_segmentation.py? createLabels.py is only able to make gtFine_labelids.png.

Secondly, what's the args.gtInstancesFile in 73 line on evaluate_instance_segmentaion.py?

Thanks very much.