Regarding error while computing the SemanticSegmentationTask with num_classes=2 at loss_ fun

[kethan52]

model =SemanticSegmentationTask(
segmentation_model="unet",
encoder_name="resnet18",
encoder_weights=None,
in_channels=10,
num_classes=2,
num_filters=64,
loss="jaccard",
ignore_zeros=False,
learning_rate=0.1,
learning_rate_schedule_patience=5,
)

trainer = pl.Trainer(

              gpus=1,
              max_epochs=10,
              precision=16,
              log_every_n_steps=1,
              max_steps=2,
           )

trainer.fit(model,dl)

---

Using 16bit native Automatic Mixed Precision (AMP)
GPU available: True, used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs
/usr/local/lib/python3.7/dist-packages/pytorch_lightning/trainer/configuration_validator.py:133: UserWarning: You defined a validation_step but have no val_dataloader. Skipping val loop.
rank_zero_warn("You defined a validation_step but have no val_dataloader. Skipping val loop.")
Missing logger folder: /content/lightning_logs
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]

| Name | Type | Params

0 | model | Unet | 14.4 M
1 | loss | JaccardLoss | 0
2 | train_metrics | MetricCollection | 0
3 | val_metrics | MetricCollection | 0
4 | test_metrics | MetricCollection | 0

14.4 M Trainable params
0 Non-trainable params
14.4 M Total params
28.701 Total estimated model params size (MB)
Epoch 0: 0%
0/4 [00:00<?, ?it/s]

ValueError Traceback (most recent call last)
in ()
21 )
22
---> 23 trainer.fit(model,dl)

36 frames
/usr/local/lib/python3.7/dist-packages/torchmetrics/utilities/checks.py in _check_shape_and_type_consistency(preds, target)
113 else:
114 raise ValueError(
--> 115 "Either preds and target both should have the (same) shape (N, ...), or target should be (N, ...)"
116 " and preds should be (N, C, ...)."
117 )

ValueError: Either preds and target both should have the (same) shape (N, ...), or target should be (N, ...) and preds should be (N, C, ...).

[kethan52]

Like as it was the binary segmentation i tried cross-entropy even for that I am getting error can anyone help to solve this problem

[calebrob6]

What dataloader/datamodule are you using? Can you post the code / initialization?

[kethan52]

class CustomDataset(RasterDataset):
filename_glob = "*ryone.tif"

def getitem(self, query):
a_max=0.1
data=super().getitem(query)
data["mask"]=transform(data)["image"][None,8,:,:]

data["mask"] = (data["mask"]<a_max)*1
data["mask"].float()

del data["bbox"]
return data

ds = CustomDataset(data_root)
sampler = RandomGeoSampler(ds, size=76.8, length=4,roi=None)
dl = DataLoader(ds, sampler=sampler, collate_fn=stack_samples)
print(dl)
print(sampler)
print(ds)

i used the 76.8 becouse each is dividing by 0.3 so i get the size as torch.Size([1, 1, 256, 256])

[calebrob6]

Can you print the shape of batch["mask"], it should be (batch size, height, width) I think. E.g.:

for batch in dl:
    print(batch["mask"].shape)
    break

[kethan52]

yes, the shape of the mask is torch.Size([1, 1, 256, 256])

for x in dl:
data = x['image']
target =x['mask']
print(data.shape)
print(target.shape)

torch.Size([1, 10, 256, 256])
torch.Size([1, 1, 256, 256])
torch.Size([1, 10, 256, 256])
torch.Size([1, 1, 256, 256])
torch.Size([1, 10, 256, 256])
torch.Size([1, 1, 256, 256])
torch.Size([1, 10, 256, 256])
torch.Size([1, 1, 256, 256])

[calebrob6]

I think you need to change your dataloader to return masks of size (batch size, height, width)

[kethan52]

Thank you soo much @calebrob6 i think it worked well

[calebrob6]

You're welcome!

[kethan52]

But can I know why my epochs was not moving?