Update README.md

simpler controls: sampling (none/uniform) drives frame rate, drop max-frames knob

live progressive preview during analysis

frame sampling mode (uniform/none) + cleaner slider labels

skinned + no-skin downloads (Mixamo-style) + frame-count slider

cover full video length (remove 120-frame cap; stride long clips)

Update README.md

fix texture V-flip on transplant

sync: texture transplant + robust node lookup + de-tremor smoothing

Upload folder using huggingface_hub

robust node/npm lookup + conversion-path logging

Gaussian de-tremor smoothing + skin skeleton root fix

transplant embedded FBX texture onto the rig GLB

temporal smoothing (median5+mean3) to reduce jitter

drop nodejs/npm from packages.txt (HF already installs Node 20 via NodeSource; avoids libnode-dev conflict)

bundle pkl+ply as zip; remove PLY viewer window

FBX->GLB via three.js (fixes broken T-pose + invalid animation); add nodejs

motion cache + raw-motion debug dump + SAM3D/rig showcase

fix: resample world_rotations with num_frames

wire SAM 3D Body -> retarget -> GLB (MHR world rotations + rest pose)

log MHR skeleton structure (names/parents/offsets)

add MoGe (FOV estimator dep used by default by SAM 3D Body)

add detectron2 transitive deps (cloudpickle, omegaconf, termcolor, tabulate, iopath, matplotlib, tqdm, pillow)

UI: SAM 3D Body only, no backend dropdown

pin torch 2.4.1 (cu12.1) to match Space CUDA 12.3; add ninja for detectron2 build

SAM 3D Body discovery deploy: torch+detectron2 stack, runner logs output structure

remove mediapipe backend; clean space

mediapipe CPU backend; hide demo+gated; python 3.11

remove __pycache__

clean pycache

vid2rig: video->rigged-model gradio tool

initial commit