RecursionError: Caught RecursionError in DataLoader for training with custom dataset

[erenuzyildirim]

Hi, I encounter the RecursionError: maximum recursion depth exceeded while calling a Python object for training with the custom dataset.

I applied the solution of setting the sys.getrecursionlimit() to a higher value but the error returns to Segmantation Fault(core dumped).

Thanks.

[albmarie]

Also having the exact same issue. Interesed if anyone has an explanation of why this is happening.

[albmarie]

I found the source of the RecursionError in my case.

In custom.py file, I changed by accident what contains prepare_train_img(self, idx) and prepare_test_img(self, idx) from: return self.pipeline(results) to self.pipeline(results) return results

I thought at first doing so was equivalent, but it is not. For example, when returning results after the call self.pipeline(results), img in results dictionary is not normalized (dtype=uint8), which is not the case with return self.pipeline(results).

It is expected that img should be normalized based on the pipeline I used:

pipeline = [ dict(type='LoadImageFromFile'), dict( type='MultiScaleFlipAug', img_scale=(2048, 1024), flip=False, transforms=[ dict(type='Resize', keep_ratio=True), dict(type='RandomFlip'), dict(type='Normalize', **img_norm_cfg), dict(type='ImageToTensor', keys=['img']), dict(type='Collect', keys=['img']), ])

More generaly, I guess the RecursionError occurs because the dataloader got an "invalid" (for whatever reason) dictionary after calling __getitem__(self, idx) class method from dataset.

Hope it helps someone.

[MeowZheng]

no response for a long time