Vision: IndexError: list index out of range

[17sarf]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

1. Run the TorchVision Object Detection Finetuning Tutorial.
2. For the model, I used the instructions for

1. Modifying the model to add a different backbone

But I keep getting the following error:

`---------------------------------------------------------------------------
IndexError Traceback (most recent call last)
in
4 for epoch in range(num_epochs):
5 # train for one epoch, printing every 10 iterations
----> 6 train_one_epoch(model, optimizer, train_loader, device, epoch, print_freq=10)
7 # update the learning rate
8 lr_scheduler.step()

/Volumes/Samsung_T5/OneDrive - Coventry University/detector/faster_rcnn_v23/engine.py in train_one_epoch(model, optimizer, data_loader, device, epoch, print_freq)
28 targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
29
---> 30 loss_dict = model(imgs1, targets)
31
32 losses = sum(loss for loss in loss_dict.values())

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torch/nn/modules/module.py in __call__(self, input, *kwargs)
530 result = self._slow_forward(input, *kwargs)
531 else:
--> 532 result = self.forward(input, *kwargs)
533 for hook in self._forward_hooks.values():
534 hook_result = hook(self, input, result)

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torchvision/models/detection/generalized_rcnn.py in forward(self, images, targets)
69 features = OrderedDict([('0', features)])
70 proposals, proposal_losses = self.rpn(images, features, targets)
---> 71 detections, detector_losses = self.roi_heads(features, proposals, images.image_sizes, targets)
72 detections = self.transform.postprocess(detections, images.image_sizes, original_image_sizes)
73

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torch/nn/modules/module.py in __call__(self, input, *kwargs)
530 result = self._slow_forward(input, *kwargs)
531 else:
--> 532 result = self.forward(input, *kwargs)
533 for hook in self._forward_hooks.values():
534 hook_result = hook(self, input, result)

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torchvision/models/detection/roi_heads.py in forward(self, features, proposals, image_shapes, targets)
754 matched_idxs = None
755
--> 756 box_features = self.box_roi_pool(features, proposals, image_shapes)
757 box_features = self.box_head(box_features)
758 class_logits, box_regression = self.box_predictor(box_features)

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torch/nn/modules/module.py in __call__(self, input, *kwargs)
530 result = self._slow_forward(input, *kwargs)
531 else:
--> 532 result = self.forward(input, *kwargs)
533 for hook in self._forward_hooks.values():
534 hook_result = hook(self, input, result)

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torchvision/ops/poolers.py in forward(self, x, boxes, image_shapes)
186 rois = self.convert_to_roi_format(boxes)
187 if self.scales is None:
--> 188 self.setup_scales(x_filtered, image_shapes)
189
190 scales = self.scales

~/opt/miniconda3/envs/torch/lib/python3.8/site-packages/torchvision/ops/poolers.py in setup_scales(self, features, image_shapes)
159 # get the levels in the feature map by leveraging the fact that the network always
160 # downsamples by a factor of 2 at each level.
--> 161 lvl_min = -torch.log2(torch.tensor(scales[0], dtype=torch.float32)).item()
162 lvl_max = -torch.log2(torch.tensor(scales[-1], dtype=torch.float32)).item()
163 self.scales = scales

IndexError: list index out of range`

I tried different models and adjusted the actors and scales, but keep getting this error.

Environment

PyTorch version: 1.4.0
Is debug build: No
CUDA used to build PyTorch: None

OS: Mac OSX 10.15.3
GCC version: Could not collect
CMake version: Could not collect

Python version: 3.8
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip] numpy==1.18.1
[pip] torch==1.4.0
[pip] torchvision==0.5.0
[conda] blas                      1.0                         mkl  
[conda] mkl                       2019.4                      233  
[conda] mkl-service               2.3.0            py38hfbe908c_0  
[conda] mkl_fft                   1.0.15           py38h5e564d8_0  
[conda] mkl_random                1.1.0            py38h6440ff4_0  
[conda] pytorch                   1.4.0                   py3.8_0    pytorch
[conda] torchvision               0.5.0                  py38_cpu    pytorch

[fmassa]

Hi,

Make sure that the feature map names returned by your backbone are compatible with the RoIPooler, see https://github.com/pytorch/vision/blob/1c7aa0c0910d518971e694620b58485704a006a8/torchvision/models/detection/faster_rcnn.py#L202

I believe there might be a mismatch there (like, strings vs ints?)

[17sarf]

@fmassa thank you very much! That seems to have resolved the issue.

[fmassa]

Cool, closing it then

[EMCP]

Had this same error occur while attempting to perform the same tutorial.. but with ResNet101 .. perhaps the docs should be updated or a note added to reflect this ?

I fixed the same error with just changing from

roi_pooler = torchvision.ops.MultiScaleRoIAlign(featmap_names=[0],
                                                        output_size=rpn_pooling_size,
                                                        sampling_ratio=2)

into

` roi_pooler = torchvision.ops.MultiScaleRoIAlign(featmap_names=["0"], output_size=rpn_pooling_size, sampling_ratio=2)

[fmassa]

@EMCP could you please send a PR fixing the tutorial?

[nvnnghia]

hi @EMCP Could you please share your fasterRCNN with resnet101 code? Thanks