# DL_python_script

**Repository Path**: zjchenchujie/DL_python_script

## Basic Information

- **Project Name**: DL_python_script
- **Description**: Some useful python scripts for data preprocess in Deep-Learning
- **Primary Language**: Python
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2019-11-14
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README


# Data preprocess utils for deep learning


- [image_xml_matchment.py]() - checking matchment of image (JPEGImages) and xml file (Annotations) in voc dataset.
image or xml files that don't have a correspond match will be moved to the directory named 'unmatched'.

- [image_area_cover.py]() - a fixed area of image will be covered with black background. If there were any targets 
included in this area, the targets won't be covered.

- [generate_train_test_filelist.py]() - generate train and test filelist in ImageSet/Main directory.
Both voc format yolo format filelist will be generated.

- [convert_voc_xml_to_yolo_label.py]() - Convert voc xml annotations to yolo text labels in directory 'labels'.

- [check_xml.py]() - Check if the boxes in xml were valid.

- [copy_files.py]() - Copy files from one directory to another.

- [eval_ap_pr.py]() - Evaluate average precision and precision.

- [extract_video_frames.py]() - Extract video frames.

- [classification_accuracy.py]() - classification accuracy

- [voc2coco.py]() - voc annotation (xml) to coco annotation (json).

- [coco2voc.py]() - coco annotation (json) to voc annotation (xml).

- [draw_anchors.py]() - Draw anchors distribution.

- [k_means_yolov3_multi.py]() - generate yolov3 anchors by k-means.

- [generate_vocxml_from_det.py]() - generate voc xml files from detection results.

- [append_object_in_xml.py]() - append object in xml.

- [delete_marked_xml_img.py]() - Delete xml marked with delete, ignore or empty.

- [draw_voc_rectangle.py]() -  Draw object rectangles on image and save this image in `img_with_rectangle` directory.

- [generate_subset.py]() - select a subset in dataset.

###data augmentation

- [crop.py]() - crop images. 
- [flip.py]() - flip images, vertically, mirror, or both.
- [resize.py]() - resize images. If fixed_ratio is True, then image should be kept original ratio with padding.
- [rotate.py]() - rotate images. 
- [sample_augmentation_mosaic.py]() - mosaic image. doesn't work well currently.