search

dvlab-research / PanopticFCN

Fully Convolutional Networks for Panoptic Segmentation (CVPR2021 Oral)
Apache License 2.0
391 stars 53 forks source link

数据集问题 #39

Closed kuozhijishou closed 2 years ago

kuozhijishou commented 2 years ago

按照指引做的cityspaces数据集 运行过程中发现：batched_input 中没有 "instances"键值，导致报错 请问是我哪里做错了吗

kuozhijishou commented 2 years ago

Traceback (most recent call last): File "/data/share/liwqg/workpath/code/python/segmentation/detectron2/detectron2/engine/train_loop.py", line 149, in train self.run_step() File "/data/share/liwqg/workpath/code/python/segmentation/detectron2/detectron2/engine/defaults.py", line 494, in run_step self._trainer.run_step() File "/data/share/liwqg/workpath/code/python/segmentation/detectron2/detectron2/engine/train_loop.py", line 273, in run_step loss_dict = self.model(data) File "/home/liwqg/miniconda3/envs/torch/lib/python3.9/site-packages/torch/nn/modules/module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "/home/liwqg/workpath/code/python/segmentation/detectron2/projects/PanopticFCN/panopticfcn/panoptic_seg.py", line 99, in forward return self.losses(pred_centers, pred_regions, pred_weights, encode_feat, gt_dict) File "/home/liwqg/workpath/code/python/segmentation/detectron2/projects/PanopticFCN/panopticfcn/panoptic_seg.py", line 142, in losses loss_pos_ths, loss_pos_sts, idx_feat_th, weighted_values, idx_feat_st, thing_nums, stuff_nums = \ ValueError: not enough values to unpack (expected 7, got 0)

yanwei-li commented 2 years ago

Hi, the cityscapes dataset should be first processed for training. Maybe the third-party implementation on Cityscapes could help you.