'NoneType' object has no attribute 'build_dataset'

[sxc95827]

Hi Burhan,

There is an issue when I run your code. It seems that yolov8 has modified some interfaces.

Traceback (most recent call last):
File "c:\Users\16119\Desktop\ClassifyDefectMap_MixedWM38-master\yolo_wafer_classification.py", line 13, in
model.trainer.build_dataset(img_path=data_dir) # loads data, but misses 'val_dir'
AttributeError: 'NoneType' object has no attribute 'build_dataset'

[Burhan-Q]

@sxc95827 yes you're right. This was a one-off project I did for an interview but I should refactor it to use a more up-to-date version of ultralytics 😆

[NG5156]

Hi Burhan,
When I calculate the confusion matrix,1 have a question,why your code is accuracy = (tp[c] + 0) / (tp[c] + 0 + fp[c] + fn[c]) # zeros are True-Negatives, but there are none for this dataset ?
The data set consists of normal samples, Why not use Accuracy =（TP+TN)/(TP+FP+TN+FN),

[Burhan-Q]

@NG5156 as I mentioned previously, this was a one-off project and needs updating/refactoring (including the calculation you mentioned). I believe this value should be extractable from the confusion matrix returned.