Weihan Xu PRO

WitneyWW

https://wx83.github.io/

wx83

AI & ML interests

None yet

Recent Activity

published a Space about 1 hour ago

WitneyWW/pyramid2-review

updated a Space about 2 hours ago

WitneyWW/pyramid2-review

published a Space about 23 hours ago

WitneyWW/ego4d-annot-review

View all activity

Organizations

None yet

WitneyWW 's Spaces 120

JavisDiT source motion extraction

🌊

Explore extracted motion fields from video clips

Input / Output Pairs — AV Object-Removal Pipeline

🎬

Remove objects from videos and isolate their sounds

Speech-free & Music-free Clips — Random Sample

🎲

Browse speech‑free and music‑free video clips

JavisDiT edit - anchor matrix (audiofix vs pre-fine-tune)

🧮

Explore video edit results with reference‑anchor comparisons

JavisDiT audio-repair finetune - step 14000 (10 test clips)

🎼

Compare video object removal edits with original and target clips

JavisDiT audio-repair - step 33500 vs pre-fix baseline

⚖

Explore audio repair results and compare before/after clips

Removing Sounding Objects — Speech-Filtered Run

🦉

Remove objects and their sound from video clips

Wan2.2-5B + dual action→motion encoders — World-model rollouts

🎛

Explore model rollouts of tactile and visual predictions

AV Curation Pipeline — 10 Successes, 5 Failures

🔇

Remove objects from video and strip their sound

JavisDiT audio - mel filterbank fix, encode/decode comparison

🔊

Compare original, fixed, and broken audio clips with metrics

JavisDiT Edit - base 0728 step 139000 (10 test clips)

🎬

Remove objects from videos and view edited vs original

Non-speech Audio+Video Removals — JAVEdit-100k & AVISet

🔇

Search video edits that remove objects and their sounds

InsAVE-80K Sample Viewer

🎬

Explore audio‑video editing samples with an interactive viewer

JavisDiT Edit — text_both_base_0728 step 104000 (20 test clips)

🎬

Add or remove objects in videos with matching audio

JavisDiT Edit — text_both_base_0728 step 85000 (full test set)

🎬

Explore AI-generated video object edits and metrics

Visuo-Tactile MoT - vmx injection rollouts

🤚

Explore ground‑truth vs prediction strips

Action to Camera Motion — a learned ActionImage

🧭

Predict dense camera motion from robot actions and visual state

Action-Image MoT World Model — Rollouts, Short vs Long, Train vs Test

🎯

Explore predicted robot vision and touch rollouts

JavisDiT Edit — text_both_base_0728 step 71000

🎬

Explore video object add/remove edits with audio

JavisDiT Edit — text_both_base_0728 step 56000

🎬

Explore AI-generated object add/remove edits in videos

History-Conditioned MoT (no diffusion forcing) — Rollouts

⏮

Explore model predictions vs ground truth

JavisDiT Edit — text_both_base_0728 step 39000

🎬

Explore AI video object add/remove with audio previews

Tactile Retrieval Oracle Upper Limit

🦷

Explore tactile prediction limits with interactive visual report

Action Images demo — pose → multi-view pixel actions

🎯

Generate multi‑view action images from robot pose data