Depth processing for videos stored as frame-by-frame images in a folder, with customizable options

[dimtsiakm]

This PR adds support for processing videos stored as frame-by-frame images in a folder, enabling depth estimation on pre-extracted frames.

Changes

• Modified run_streaming_frames.py to read frames from a folder instead of a video file
• Automatically detects and processes all .jpg files in the input directory
• Saves individual depth frames as colored JPG images with the same inferno colormap used in the output video
• Maintains all existing functionality (stride, max_len, fps control, grayscale mode)
• Uses default 30 FPS for folder-based processing (overridable with --target_fps)

Usage

python [run_streaming_frames.py] --input_video /path/to/frames/folder --output_dir ./outputs/ --max_len 5000