Hello,
I would like to run detectron on my own dataset.
I just have to detect and segment a single class. In order to do that I converted my dataset to coco format as follows:
I made a folder containing all the train pics in data/coco/images/satellite
Every pic has a random string as a name.
I generated the annotations for each instance and I saved the json file in data/coco/annotations
Every instance has these information:
{'id': 955, 'image_id': 'haF473KL7', 'category_id': 1, 'iscrowd': 0, 'area': 146, 'bbox': [446.0, 721.0, 19.0, 13.0], 'segmentation': [[449.0, 733.5, 446.0, 725.5, 462.5, 721.0, 464.0, 728.5, 449.0, 733.5]], 'width': 512, 'height': 512}
id : is just a counter that counts the instances
image_id : contains the file name of the image (without format)
I also changed the dataset catalog and the train_net_step.py file accordingly.
When I run the training the console returns this error:
INFO roidb.py: 50: Appending horizontally-flipped training examples...
INFO roidb.py: 52: Loaded dataset: satellite_train
INFO roidb.py: 143: Filtered 208138 roidb entries: 208138 -> 0
INFO roidb.py: 69: Computing image aspect ratios and ordering the ratios...
INFO roidb.py: 71: done
INFO roidb.py: 75: Computing bounding-box regression targets...
INFO roidb.py: 77: done
Traceback (most recent call last):
File "tools/train_net_step.py", line 457, in
main()
File "tools/train_net_step.py", line 233, in main
cfg.TRAIN.DATASETS, cfg.TRAIN.PROPOSAL_FILES)
File "/home/stefan/Satellite/mask-rcnn.pytorch/lib/datasets/roidb.py", line 79, in combined_roidb_for_training
_compute_and_log_stats(roidb)
File "/home/stefan/Satellite/mask-rcnn.pytorch/lib/datasets/roidb.py", line 230, in _compute_and_log_stats
classes = roidb[0]['dataset'].classes
IndexError: list index out of range
Hello, I would like to run detectron on my own dataset. I just have to detect and segment a single class. In order to do that I converted my dataset to coco format as follows:
I made a folder containing all the train pics in data/coco/images/satellite Every pic has a random string as a name.
I generated the annotations for each instance and I saved the json file in data/coco/annotations Every instance has these information:
{'id': 955, 'image_id': 'haF473KL7', 'category_id': 1, 'iscrowd': 0, 'area': 146, 'bbox': [446.0, 721.0, 19.0, 13.0], 'segmentation': [[449.0, 733.5, 446.0, 725.5, 462.5, 721.0, 464.0, 728.5, 449.0, 733.5]], 'width': 512, 'height': 512}
id : is just a counter that counts the instances image_id : contains the file name of the image (without format)I also changed the dataset catalog and the train_net_step.py file accordingly. When I run the training the console returns this error:
Any idea how to solve this issue? Thank you