VallueError: Found input variables with inconsistent numbers of samples

[iwillflyandfly]

When I try to evaluating a pretrained PatchCore model like cable, it raises ValueError :" Found input variables with inconsistent numbers of samples: [15360000, 7526400]" INFO:main:Evaluating dataset [mvtec_cable] (1/1)... INFO:patchcore.patchcore:Loading and initializing PatchCore. INFO:main:Embedding test data with models (1/1) INFO:main:Computing evaluation metrics. Traceback (most recent call last): File "src/load_and_evaluate_patchcore.py", line 286, in main() File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/click/core.py", line 1130, in call return self.main(args, kwargs) File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/click/core.py", line 1055, in main rv = self.invoke(ctx) File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/click/core.py", line 1689, in invoke return _process_result(rv) File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/click/core.py", line 1626, in _process_result value = ctx.invoke(self._result_callback, value, ctx.params) File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/click/core.py", line 760, in invoke return __callback(args, **kwargs) File "src/load_and_evaluate_patchcore.py", line 154, in run pixel_scores = patchcore.metrics.compute_pixelwise_retrieval_metrics( File "/home/lihang/patchcore/patchcore-inspection/src/patchcore/metrics.py", line 47, in compute_pixelwise_retrieval_metrics fpr, tpr, thresholds = metrics.roc_curve( File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/sklearn/metrics/_ranking.py", line 979, in roc_curve fps, tps, thresholds = _binary_clf_curve( File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/sklearn/metrics/_ranking.py", line 740, in _binary_clf_curve check_consistent_length(y_true, y_score, sample_weight) File "/home/lihang/anaconda3/envs/lihang-conda/lib/python3.8/site-packages/sklearn/utils/validation.py", line 387, in check_consistent_length raise ValueError( ValueError: Found input variables with inconsistent numbers of samples: [15360000, 7526400]

[Confusezius]

This seems to be pointing to the issue that either the generated anomaly segmentation maps OR the ground truth masks are of different sizes, since compute_pixelwise_retrieval_metrics flattens the generated anomaly segmentations and the ground truth masks for evaluation.

Can you check this?

[iwillflyandfly]

Many thanks!

[TheWangYang]

I also encountered this problem during the evaluation. I would like to know how to solve this problem.

[TheWangYang]

I finally solved the problem and the analysis is as follows:

• When training my model input parameters are set as follows: --resize 256 --imagesize 244;
• But when I loaded the pre-trained model, I set the model input parameters as follows: --resize 366 --imagesize 320;
• The reason is that when training and evaluating the same model, the image input sizes are not consistent;
• Finally, modify the input parameters of the evaluation model to those input during model training;