Unsatisfactory video results

[Condor9]

Hi, I run the code based on your default arguments, but the result of the generated video is almost completely black, with unclear target objects in just a few frames. What's the reason for this? The result of running the code is as follows:

loading scene params from: C:\PhysDreamer-main\projects\inference\configs\carnation.py => loading camera from blender format transforms_train.json => loading camera from blender format transforms_train.json len of test dataset 209 load gaussians from: data/physics_dreamer/carnations/point_cloud.ply ... num gaussians: 1037279 loaded 130812 clean points from: data/physics_dreamer/carnations\clean_object_points.ply init young modulus: 2140628.25 poisson ratio: 0.264644056558609 scale, shift tensor(2.8624, device='cuda:0') tensor(0.7803, device='cuda:0') simulation aabb: tensor([[0.2611, 0.5404, 0.1217], [0.3752, 0.7084, 0.3275]], device='cuda:0') Downsample with ratio: 0.1 => staring downsample with kmeans from 133564 points to 13356 points => downsample with kmeans takes 159.35061621665955 seconds Downsampled to: 13356 by 0.1 Number of non-empty voxels: 99 Max points in voxel: 841.0 Min points in voxel: 0.0 Avg points in voxel: 134.9090909090909 Warp 0.10.1 initialized: CUDA Toolkit: 11.5, Driver: 12.5 Devices: "cpu" | Intel64 Family 6 Model 158 Stepping 13, GenuineIntel "cuda:0" | NVIDIA GeForce RTX 2080 Ti (sm_75) Kernel cache: C:\Users\AppData\Local\NVIDIA Corporation\warp\Cache\0.10.1 (13356,) torch.Size([13356]) Module physdreamer.warp_mpm.mpm_data_structure load on device 'cuda:0' took 4.20 ms Particles initialized from torch data. Total particles: 13356 Module mpm_data_structure load on device 'cuda:0' took 3.63 ms number of freeze grid 24 num freeze pts in total 1835 num moving pts 11521 Module mpm_utils load on device 'cuda:0' took 244.40 ms => zero init the last layer for Spatial MLP => zero init the last layer for Spatial MLP => loading: models/physdreamer/carnations/model\velo_fields.pt => loading: models/physdreamer/carnations/model\sim_fields.pt torch.Size([13356]) 6986 apply force mask 0%| | 0/90 [00:00<?, ?it/s]Module physdreamer.warp_mpm.mpm_solver_diff load on device 'cuda:0' took 2.66 ms 100%|██████████| 90/90 [04:17<00:00, 2.86s/it] number of simulated frames: 91 num camera viewpoints: 91 save video to output/carnations/demos\inference_demo_carnation_sv_gres64_substep768_model_0_force_1_mag_2.0_point_0_movingcamera_camid_0_inference_demo_carnation_sv_gres64_substep768_model_0_force_1_mag_2.0_point_0.mp4

[a1600012888]

Can you upload the video here, then I can try to debug.

[Condor9]

Thanks for your reply, it has been resolved!