Skip to content

Latest commit

 

History

History
156 lines (137 loc) · 5.62 KB

DATA_README.md

File metadata and controls

156 lines (137 loc) · 5.62 KB

Data Preperation

We provide cloud links for downloading the datasets used in this project, along with their directory structure and code.
For the IITB Corridor dataset, due to its large size, we provide data preprocessing code instead.
We express our sincere gratitude to researchers Cewu Lu [Avenue], Weixin Luo & Wen Liu [SHTech], and Royston Rodrigues [Corridor] for providing their datasets.

1. CUHK Avenue

The dataset can be downloaded from the official website [Link]. However, there might be some differences in the directory structure compared to the dataset used in this project. You can set up our environment using the download link provided below.

File Name Download
avenue.zip Google Drive 
avenue
├── testing
    └── frames
        └── 01
            ├── 0000.jpg
            ├── ...
            └── 1438.jpg
        ├── 02
        ├── ...
        └── 21
└── training
    └── frames
        ├── 01
            ├── 0000.jpg
            ├── ...
            └── 1363.jpg
        ├── 02
        ├── ...
        └── 16
├── avenue.mat

2. ShanghaiTech Campus

The dataset can be downloaded from the official website [Link]. However, there might be some differences in the directory structure compared to the dataset used in this project. You can set up our environment using the download link provided below.

File Name Download
shanghai.vol1.egg Naver Mybox
shanghai.vol2.egg Naver Mybox
shanghai.vol3.egg Naver Mybox
shanghai.vol4.egg Naver Mybox
shanghai.vol5.egg Naver Mybox
shanghai.vol6.egg Naver Mybox
shanghai.vol7.egg Naver Mybox
shanghai.vol8.egg Google Drive 
shanghai
├── testing
    └── 01_0014
        ├── 000.jpg
        ├── ...
        └── 264.jpg
    ├── 01_0015
    ├── ...
    └── 12_0175
├── training
    └── 01_001
        ├── 0.jpg
        ├── ...
        └── 763.jpg
    ├── 01_002
    ├── ...
    └── 13_007
├── testframemask
    ├── 01_0014.npy
    ├── ...
    └── 12_0175.npy

3. IITB Corridor

The dataset can be downloaded from the official website [Link]. However, there might be some differences in the directory structure compared to the dataset used in this project. Therefore, please restructure the directories according to the structure below before running the experiments.

iitb
├── testing
    └── 000209
        ├── 0000.jpg
        ├── ...
        └── 0547.jpg
    ├── 000210
    ├── ...
    └── 000358
├── training
    └── 000001
        ├── 0000.jpg
        ├── ...
        └── 0341.jpg
    ├── 000002
    ├── ...
    └── 000208
├── groundtruth
    └── 000209
        └── 000209.npy
    ├── 000210
    ├── ...
    └── 000358

We downloaded the dataset from the official website and executed the following code in the Test_IITB-Corridor directory. Below is the code we created to convert video into frames and save them. Once the frame conversion is complete, you can restructure the directories according to the above structure. The training dataset has been processed in the same way.

import sys
sys.path.append('.')

import cv2
import os
import glob
import re
from fastprogress import progress_bar


def extract_numbers(file_name):
    numbers = re.findall(r'(\d+)', file_name)
    return tuple(map(int, numbers))

#==========(Test/Train)==========#
work='Test'
#================================#

video_list = glob.glob(f'./{work}/*')
video_list = sorted(video_list, key=extract_numbers)
print(video_list)  # ['./Test/000209', './Test/000210', './Test/000211', ...]

video_name_list = os.listdir(f'./{work}')
video_name_list = sorted(video_name_list, key=extract_numbers)
print(video_name_list) # ['000209', '000210', '000211', ...]

for i, video in progress_bar(enumerate(video_list), total=len(video_list)):
    video_name = video_name_list[i]
    video += f'/{video_name}.avi'

    print(f'--------------{i, video_name}------------------')
    if not os.path.exists(f"./{work}_fr/{video_name}"):
        os.makedirs(f"./{work}_fr/{video_name}")

    vidcap = cv2.VideoCapture(video)
    success,image = vidcap.read()
    count = 0
    while success:
      cv2.imwrite(f"./{work}_fr/{video_name}/%04d.jpg" % count, image)     # save frame as JPEG file
      success,image = vidcap.read()
      count += 1

    print("finish! convert video to frame")