Data Crawling

Data Crawling

Search

Path: toolset/crawl/search/search.sh

Description: This script uses the YouTube API to search for 360-degree videos based on a predefined list of keywords and suffix terms. Note that the retrieved videos are confirmed as 360-degree videos but do not guarantee support for Spatial Audio. To filter videos that support Spatial Audio, use a downloader (e.g., yt-dlp with an audio channel count filter).

Usage Instructions:

Prepare a keyword list file (one keyword per line).
Modify the parameters in search.sh, such as keyword_file, postfix(i.e. the qualifying term), and output_dir. Refer to the code comments for detailed parameter descriptions.
Run bash search.sh to execute the search. Results are saved in CSV format under the output_dir directory, organized by keyword.

Channel-Based Crawling

Analyze channels based on search results

Path: toolset/crawl/channel/channel_analyzer.py

Description: Analyzes frequently appearing channels based on search results from the search module. Outputs in CSV format.

Usage Instructions:

python channel_analyzer.py -i [input-dir] -o [output-file]

Get Video IDs for channels

Path: toolset/crawl/channel/get_channel_vids.py

Description: Retrieves all 360-degree video IDs from the most promising channels (those containing the most search results). The output directory contains CSV files named by Channel ID, each containing all 360-degree video IDs for the corresponding channel.

Usage Instructions:

python get_channel_vids.py -i [input-csv] -o [out-dir] -t [threshold]

Where threshold specifies the minimum number of times a channel must appear in search results to be retained.

Get test video metadata

Path: toolset/crawl/channel/get_test_list.py

Description: For each channel, randomly samples several video segments from the Channel ID's video ID information, and outputs in CSV format recognizable by download scripts.

Usage Instructions:

python get_test_list.py -i [in-dir] -o [out-dir] [-n [sample-size]]

Here in-dir contains the CSV files named by Channel ID obtained from the previous module, and out-dir will contain CSV files named by each Channel ID, with each file containing up to 10 Video IDs for the corresponding channel.

Download test videos & Check

Use your preferred downloader or scripts from the Download section to download test segments, then perform manual verification or use the cleaning pipeline from the Data Cleaning section to verify channel quality, filtering out usable and unusable Channel IDs (both can be included in the Channel Black List).

Get Video IDs based on Channel IDs

Path: toolset/crawl/channel/get_channels_vids.py

Description: Retrieves video IDs to be downloaded based on a list of available Channel IDs. All resulting Video IDs are guaranteed to be 360-degree videos. Due to API limitations, Spatial Audio support is verified during the download phase rather than this stage.

Usage Instructions:

python get_channels_vids.py -i [input-csv] -o [output-csv]

Download

Use your preferred downloader or scripts from the Download section for video downloading.

Video-Based Crawling

Filter by Blacklist

Path: toolset/crawl/filter/filter_exist.py

Description:

Utilizes the established video and channel blacklists from Channel-Based Crawling to filter out confirmed unnecessary video entries from search results, generating a smaller-scale list for manual review.

Usage Instructions:

Prepare the Video ID blacklist database and Channel ID blacklist database (including all manually verified usable/unusable video or channel IDs). Modify video_db_path and channel_db_path in the script.
Update the search results path folder_path (output from the Search module) and output file path output_csv in the script.
Run python filter_exist.py to execute the filtering.

Download Test Videos & Verification & Full Download

Use a downloader of choice to fetch video clips for screening. After verification, proceed with full downloads of usable videos.

Download

Full Video Download

Path: toolset/crawl/download/download_list.sh

Description: Downloads all videos based on Video IDs provided in a CSV file. Supports multi-process downloading. Results are logged in success_list.txt and fail_list.txt in the current directory.

Usage Instructions:

Modify relevant parameters in download_list.sh, then execute:

bash download_list.sh

Batch Download Function

The download_list_360 function in toolset/crawl/core/download/download_list.py enables batch downloading. Refer to the function documentation for usage. Common configurations include:

Clip Downloading: Use specify_start and time_interval parameters (see function docs for details).
Multi-process Downloading: Set the number of parallel downloads via the jobs parameter.
Custom Cookies: Specify cookies using the cookie parameter.
Proxy Configuration: Configure proxy servers via the proxy parameter.