Spaces:
Sleeping
Sleeping
File size: 1,937 Bytes
6557c00 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 | ---
title: Text-to-HDR User Study
emoji: 🎬
colorFrom: indigo
colorTo: red
sdk: gradio
sdk_version: 6.0.0
app_file: app.py
pinned: false
license: cc-by-4.0
---
# Text-to-HDR — Pairwise User Study
Pairwise human comparison of three text-to-HDR methods (HDR-LTX,
X2HDR, LEDiff) on 30 cinematic prompts. Raters see 90 stacked
3-EV bracket pairs and pick which row looks more natural / more like
a real photograph.
## Files in this Space
- `app.py` — Gradio app
- `pairs.json` — 90 pair definitions (with hidden top/bottom assignment)
- `prompts.json` — 30 source prompts
- `pairs/pair_NNN.png` — 90 stacked-bracket comparison images
- `requirements.txt` — `gradio`, `huggingface_hub`
## Vote storage
Each completed rater session writes one JSONL file
(`votes/votes_<rater_id>.jsonl`) to a private HF dataset repo
(`HF_DATASET_REPO` env var). One line per pair, with the chosen label
and the recorded top/bottom method assignment.
## Local dev
```bash
pip install -r requirements.txt
python app.py
```
When `HF_TOKEN` / `HF_DATASET_REPO` env vars are not set, the app
runs locally and stores votes in `votes_<rater_id>.jsonl` next to
`app.py` (no upload).
## Deploy to a free HF Space
```bash
# 1. create the space + dataset on huggingface.co (or via the CLI)
# 2. set Space secrets:
# HF_TOKEN — write-scoped token for the dataset
# HF_DATASET_REPO — e.g. "naomi/t2hdr-user-study-votes"
# 3. push the contents of this directory to the Space repo:
huggingface-cli login
git lfs install
git clone https://huggingface.co/spaces/<your-username>/t2hdr-user-study
cp -r * /path/to/cloned/space/
cd /path/to/cloned/space
git lfs track "pairs/*.png"
git add . && git commit -m "Initial study upload" && git push
```
## Scoring
After the study closes, run `score_study.py` (TODO) to download all
JSONL files from the dataset repo, compute Thurstone Case V scores
per method, and write `results.json`.
|