dataset issues

[Hukiwi]

Thank you for sharing this repo! I'm trying to run the train.py script, but I get the following error:

dataset.py", line 116, in self.kitti_datalist = [pykitti.odometry(basedir,seq,frames=frame) for seq,frame in zip(seqs,frame_list)]
File "/root/anaconda3/envs/yolov5/lib/python3.8/site-packages/pykitti/odometry.py", line 35, in init self._load_poses() File "/root/anaconda3/envs/yolov5/lib/python3.8/site-packages/pykitti/odometry.py", line 220, in _load_poses lines = [lines[i] for i in self.frames] File "/root/anaconda3/envs/yolov5/lib/python3.8/site-packages/pykitti/odometry.py", line 220, in lines = [lines[i] for i in self.frames] IndexError: list index out of range

How can I solve it?

[gitouni]

I think the problem lies in that some datasets you designated in the config file are not downloaded in the directory. If you double check the datasets and still have problems, show your directory tree and config file here.

[Hukiwi]

The datasets designated in the config file are all downloaded in the directory. and the contents of the directory tree and config file are shown below: CalibNet_pytorch/ --|data/ --|poses/ --|00.txt --|01.txt --... --|sequences/ --|00/ --|image_2/ --|velodyne/ --|calib.txt --|times.txt --|01/ --|02/ --... --...

dataset: train: [0,1,2,3,4,5,6,7] val: [8,9,10] test: [11,12,13] cam_id: 2
voxel_size: 0.1 pooling: 3
extend_ratio: [2.5,2.5]

[gitouni]

set train set to「0」and val to 「1」and Try again. Remember to Delete the cache file“data_len.json" as well.

[Hukiwi]

set train to「0」and val to 「1」, the data reading works fine, is there something wrong with the file I downloaded?

[gitouni]

set train to「0」and val to 「1」, the data reading works fine, is there something wrong with the file I downloaded?

I think so. You can debug the python script and check the consistency between the variable "frame" and the actual number of files in your directory. Furthermore, you can set other sequence numbers for "train" and see which one causes error.

[gitouni]

Sorry, someone raised another issue to remind me that the update on dataset.py causes this problem. I changed it to the old version and you can download it again and have a try.