search

ristea / aed-mae

Other
19 stars 1 forks source link

Cannot Open data files after processing #8

Closed JPerAsperaadAstra closed 1 month ago

JPerAsperaadAstra commented 1 month ago

Hi, dear authors. Thanks for your outstanding work. And here's a problem I encountered after processing the dataset following README: I created anomaly sequces from UBnormal dataset and integrated them with util/create_anomalies/create_abnormal_sequences.py and util/create_anomalies/integrate_abnormal_objects.py, then I organited the dataset as following structure: -Avenue --testing ---frames ---gradients --traing ---frames ---frames_abnormal ---gradients ---masks_abnormal where frames_abnormal and masks_abnormal were generated with pre-processing scripts util/create_anomalies/main.py, gradients was generated by extractgradients.py, and frames was the original frames from avenue dataset. However, when I ran main.py on the avenue dataset, I got following errors: [ WARN:0@14.611] global loadsave.cpp:248 findDecoder imread('datasets/Avenue/training/frames_abnormal/07/541.jpg'): can't open/read file: check file path/integrity

It seems to be a data format or integrity problem, but I was able to open this jpg just fine. Could you please give me some suggestion about fixing this error and make it work? Thx

CroitoruAlin commented 1 month ago

Hello! This looks like something related to your environment. I see that you are using a relative path, please make sure that it is correct.

JPerAsperaadAstra commented 1 month ago

pleted May 26, 2024

Hi, thx for your quick response. I tried to use the absolute path but resulted in the same error. Could it be possible that my extracted frames of avenue dataset is .jpg while abnormal sequences are saved as .png? (since frames_abnormal are abnormal_sequences added to normal data)

flww213 commented 1 month ago

Image name is wrong, you can check the datasets/Avenue/training/frames_abnormal / 07, I guess the image under the path should be "0541.jpg". Therefore, modify the parameters such as len_frame_no of the getitem function in data/train_dataset.py .Or modify the image format for masks_abnormal in about 48 lines.

JPerAsperaadAstra commented 1 month ago

@flww213 Thank you so much and your suggestion is right!