README.md

metadata

title: Trajectory Endpoint Labeler
emoji: 🎯
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.1.0
python_version: 3.12
app_file: app.py
pinned: false
short_description: Label trajectory success cutoffs for Robometer data

Trajectory Endpoint Labeler

Gradio tool for labeling where robot trajectories reach task success — used to derive per-dataset success cutoff percentages for RewardFM / Robometer training.

Features

• Load trajectories from any HuggingFace dataset
• Filter by success / failure / all
• Frame-precise end-point marking with percent-of-trajectory
• Pattern analysis across labeled trajectories (mean, std, suggested cutoff)
• CSV export (labels.csv)

Usage

1. Enter HF dataset repo (e.g. jesbu1/epic_rfm)
2. Set sample counts and quality filter
3. Scrub to the success frame → save label
4. After labeling a batch, run Analyze Pattern for suggested cutoff %

Output

dataset_repo,config_name,trajectory_id,is_robot,quality_label,task,manual_end_frame,manual_end_percent,notes

Suggested cutoffs feed into dataset_success_cutoff.txt for training.

---

Part of Robometer · Built with Gradio + HuggingFace Datasets