Hello, I’m trying to train your code, but I’ve encountered an issue. When I use the pre-processed ACDC dataset provided in the project, the training works very well. However, when I use the ACDC dataset processed by myself using the code in the project, the training often stops converging around 1800 iterations, and the performance at that point is approximately mean_dice: 0.445736. Additionally, both scribble_generator.py and acdc_data_processing.py throw errors when processing the ACDC dataset. Do you have a more detailed dataset processing code that could help me achieve good results with my own processed ACDC dataset?
Hello, I’m trying to train your code, but I’ve encountered an issue. When I use the pre-processed ACDC dataset provided in the project, the training works very well. However, when I use the ACDC dataset processed by myself using the code in the project, the training often stops converging around 1800 iterations, and the performance at that point is approximately mean_dice: 0.445736. Additionally, both scribble_generator.py and acdc_data_processing.py throw errors when processing the ACDC dataset. Do you have a more detailed dataset processing code that could help me achieve good results with my own processed ACDC dataset?