vr-hmr / diagnose_output.log
zirobtc's picture
Upload folder using huggingface_hub
7e120dd
[01/08 05:07:54][INFO] [PL-Trainer] Loading ckpt: e004-s000005.ckpt
[01/08 05:07:54][INFO] [PL-Trainer] Loading ckpt: e004-s000005.ckpt
=== Loading: 0_biboo_birthday_speech.pt ===
Loading GENMO model configuration...
Using checkpoint: e004-s000005.ckpt
Loading pretrained GENMO model...
Gen only test timestep respacing: 50
Running model inference...
Preproc taken: 0.08589911460876465
Demo taken: 3.6113734245300293
============================================================
FRAME 0 COMPARISON: PREDICTION vs GROUND TRUTH
============================================================
--- In-Camera Global Orientation ---
GT (YXZ deg): yaw=-119.65, pitch= -0.06, roll=-178.54
Pred (YXZ deg): yaw=-133.52, pitch= -12.25, roll= 169.91
Diff (deg): yaw= -13.87, pitch= -12.19, roll= 348.45
--- World/Global Orientation ---
GT (YXZ deg): yaw= 169.81, pitch= 0.18, roll= 1.24
Pred (YXZ deg): yaw= 134.80, pitch= 4.41, roll= -3.63
Diff (deg): yaw= -35.01, pitch= 4.23, roll= -4.87
--- In-Camera Translation ---
GT: [ -0.045, 0.911, 1.153]
Pred: [ -0.023, 0.821, 1.147]
Diff: [ 0.022, -0.090, -0.005]
--- World Translation ---
GT: [ 0.000, 0.000, 0.000]
Pred: [ -0.001, 1.237, 0.003]
Diff: [ -0.001, 1.237, 0.003]
--- Body Pose ---
Max difference: 101.85°
Mean difference: 8.71°
============================================================
SUMMARY ACROSS ALL FRAMES
============================================================
In-camera orientation errors (mean ± std):
Yaw: 11.03° ± 1.44°
Pitch: 12.49° ± 1.06°
Roll: 348.56° ± 0.37°
World orientation errors (mean ± std):
Yaw: 37.47° ± 1.39°
Pitch: 4.45° ± 0.55°
Roll: 4.69° ± 0.32°
In-camera translation error: 0.0987m ± 0.0070m
World translation error: 1.2386m ± 0.0017m
Body pose error: 12.39° ± 22.74° (max: 171.51°)
============================================================