Error adding images to training

[SharhadBashar]

I want to modify the images used for training: i want to add my own images. how do i go around doing that? here is what i did:

1. I added the raw images to the JPEGImages folder
2. I added the segmented versions to the SegmentationClassAug folder
3. I added the image paths to train_aug.txt in voc12
4. I added their labels to train_label.npy but not sure I added them correctly
5. I added their image name and class label to cls_labels.npy in the same format as the other entires
6. did not need to add them to 20_class_labels.npy, as they were already there I added about 1500 images, extract features extracts all the 12000+ images, but when i run create pseudo label, in the line instead of printing (12000, 200) (12000, 20) it prints (18000, 200) (18000, 20) What am i doing wrong, or am i missing any steps?

Heres the error:

18605 18605
18605 18605 (16458, 20)
Traceback (most recent call last):
  File "create_pseudo_label.py", line 202, in <module>
    train_filename_list, train_label_200, train_label_20 = create_train_data(merge_filename_list, new_label_list, keep_idx_list)
  File "create_pseudo_label.py", line 126, in create_train_data
    train_label_20.append(label_20[idx])
IndexError: index 16460 is out of bounds for axis 0 with size 16458
k_cluster: 10

[WeiChihChern]

6. did not need to add them to 20_class_labels.npy, as they were already there I added about 1500 images, extract features extracts all the 12000+ images, but when i run create pseudo label, in the line instead of printing (12000, 200) (12000, 20) it prints (18000, 200) (18000, 20)

Do you mean you didn't change the '20_class_labels.npy' at all? If you include your own images, you should add your labels to '20_class_labels.npy'. If your images contain one class per image, then you can simply append your labels to the 20_class_labels.npy. If not... well it's going to be complicated.

I think the create_pseudo_label.py should generate the label of '20_class_labels.npy' automatically because the new label length is based on the processed results from the functions of def make_filename_class_dict, def merge_filename_class_dict, def generate_repeat_list, and def remove_duplicate_label, but somehow it does not.

[WeiChihChern]

Let's discuss here so others who have similar issues can refer to this.

I want to include the validation images in training as well. Thats why i dont need to make any changes to 20_class_labels.npy But I am not sure what to do with train_label.npy. These images are already included in cls_labels.npy

Any ideas what i should do next? I added the images to train_aug.txt, but dont know what to do from there

Took a glance at the training code, it does not contain a validation set. You can repeat the process of what the authors do for training set: feature extraction and pseudo label generation, and apply them for your validation set. Then load it in the training code. If it's a validation set, you would probably not want to mix it with the training set.

I am not the author(s), so my interpretation could be incorrect!

[SharhadBashar]

ok, so i found the cause of my issue 20_class_labels.npy has a size of 16458, whereas adding new images makes keep_idx_list go over that (18605), since there are more images so my question is, how is this file 20_class_labels.npy created?