Skip to content

Latest commit

 

History

History
65 lines (48 loc) · 5.24 KB

README.md

File metadata and controls

65 lines (48 loc) · 5.24 KB
Raw

Animal-Identification-from-Video

Repository of annotated videos of multiple animals and MATLAB code for processing the data

ExampleAnnotation

1. Videos

The original videos included in this repository have been sourced from Pixabay under Pixabay License

  • Free for commercial use
  • No attribution required

The video data is summarised below:

Short Name Video Name # Frames Size # Bounding boxes # Identities
Pigs Pigs_49651_960_540_500f.mp4 500 ( 960, 540) 6184 26
Koi fish Koi_5652_952_540.mp4 536 ( 952, 540) 1635 9
Pigeons (curb) Pigeons_8234_1280_720.mp4 443 (1280, 720) 4700 16
Pigeons (ground) Pigeons_4927_960_540_600f.mp4 600 ( 960, 540) 3079 17
Pigeons (square) Pigeons_29033_960_540_300f.mp4 300 ( 960, 540) 4892 28

Annotated videos are available here:

2. Data

The data is stored in Excel files with names BB_XXX.csv, where XXX stands for the name of the video. Each Excel file is organised as follows:

Name x y w h filename max_x max_y
Mahrez 1059 85 221 312 scene00001.jpg 1280 720
Torres 686 174 367 342 scene00001.jpg 1280 720
Sterling 564 132 283 145 scene00001.jpg 1280 720
Silva 102 557 356 163 scene00001.jpg 1280 720
... ... ... ... ... ... ... ...

The columns are: Names of the individual animals (class labels), the x and y coordinates in pixels of the top left corner of the bounding box, the (w)idth and (h)eight of the bounding box in pixels, the filename containing the frame of the video and the horizontal and vertical image dimensions. The file format was chosen to match that outputted by the labelling software provided by https://www.makesense.ai/.

3. Code

MATLAB functions

  • trim_video(video_name, output_file_name, number_of_frames) Trims a video to a desired number of frames.
  • store_frames(im_folder,video_file) Breaks a video into frames and saves them in a given folder. The frames are named scene00001.jpg, scene00002.jpg, etc.
  • create_annotated_video(image_folder, bb_file_name, label_flag) Creates an MP4 annotated video from the video stored as frames scene00001.jpg, scene00002.jpg, ... and a csv file with bounding boxes in the format explained in Section 2. The label flag determines whether the bounding boxes should be labelled. The default value of label_flag is True.
  • create_clips(folder,bb_file) Saves the clips from the frames in 'folder'. Each frame is examined in turn and the clips from the bounding boxes are cropped and stored in sub-folders with the name of the individual. The clip file names are formed as 'IndividualsName_frame_XXXXX.jpg'.
  • show_annotated_frame(im_folder,f_name, bb_file) Reads image 'f_name' from 'im_folder' and displays the image with overlayed labelled bounding boxes. Annotations are in the bb_file csv file (see create_annotated_video for the csv file format).

Python scripts

  • ClipClassification.ipynb Runs a classification experiment with simple feature spaces (handcrafted features).
  • SimpleFeaturesClassification.ipynb Runs a classification experiment with deep learning approaches (CNN and transfer learning).
  • CheckPythonFeatureExtraction.ipynb Shows examples of feature selection from an image.
  • FeatureExtractor.py Contains a collection of feature extraction functions.