matiss/froknows-photo-critiques
FroKnows Photo Critiques 141 manually selected, cropped screen captures of photographs shown during Jared Polin's (FroKnowsPhoto) YouTube photo-critique videos — one definitive frame per segment — each paired with the transcript commentary in which that photo is critiqued. Built for LLM evaluation of photo-editing feedback: given a frame from an editing/critique session (the photo itself or a screen shot), what does the expert commentary say about it? Contents 14… See the full description on the dataset page: https://huggingface.co/datasets/matiss/froknows-photo-critiques.
FroKnows Photo Critiques
141 manually selected, cropped screen captures of photographs shown during Jared Polin's (FroKnowsPhoto) YouTube photo-critique videos — one definitive frame per segment — each paired with the transcript commentary in which that photo is critiqued.
Built for LLM evaluation of photo-editing feedback: given a frame from an editing/critique session (the photo itself or a screen shot), what does the expert commentary say about it?
Contents
14 videos × 9–11 photos = 141 rows (single train split).
Fields
Provenance
- Source videos: Jared Polin (FroKnowsPhoto) YouTube playlist, downloaded with yt-dlp.
- Frames: manual segmentation + automated frame extraction; one definitive frame per segment hand-selected and cropped (Aug 2026).
- Commentary: Whisper ASR transcript of the audio, aligned to photo segments.
License
Frames and transcripts are derived from copyrighted YouTube content; provided for research/evaluation use. Rights in the original videos remain with the content creator.
