Video Analysis Pipeline

This repository provides the first step in behavior video QC pipeline to process and analyze videos using the VideoLoader class. The script loads video frames, extracts metadata and timestamps, and saves video frames and metadata in as a zarr group to a specified output directory.

Features

• Processes only videos containing a specified tag (default: "face").
• Loads video frames and extracts metadata.
• Supports grayscale processing.
• Saves results in an output directory.
• Tracks processing time.

Prerequisites

Ensure you have the following dependencies installed:

pip install tqdm zarr dask

Additionally, make sure utils.py and VideoLoader.py are available in your project.

Parameters

Parameter	Description	Default Value
DATA_PATH	Path to the directory containing videos.	/root/capsule/data
OUTPUT_PATH	Path to save processed video results.	/root/capsule/results
tag	Filter for video files containing this tag.	'face'
subselect	Optional filter for selecting specific videos.	'multiplane'

How It Works

1. The script scans the DATA_PATH directory for videos matching the subselect filter.
2. It initializes a VideoLoader object for each video containing the tag in its filename.
3. It processes the video in grayscale and saves results to OUTPUT_PATH.
4. It prints the processing time upon completion.
5. The output is saved as a zarr group with fields "data" and "metadata".

Example Output

Processing /root/capsule/data/video1_face.mp4
Processing /root/capsule/data/video2_face.mp4
processing only face videos for now
Total time taken: 15.32 seconds