4D-Humans

4D-Humans copied to clipboard

I have two questions. 1. In the S.3. Training details of the Supp Mat, you say that `for the HMR 2.0a variant, the ViT encoder remains frozen`, but in the...

XiaobenLi00

Backbone gpu memory increasing after each batch

8

Hello, first of all, thank you very much for sharing the implementation! I was wondering if you experienced issues with GPU memory increasing after every batch using the ViT backbone?...

DavidBoja

Hello author, may I ask if the training parameters for the original HMR model in Table 6 (B1) are consistent with those of HMR2.0b

ljqnb

Thanks for your great work, Could you share the script to evaluate the tracking performance on Posetrack dataset (HOTA, IDs, MOTA, and IDF1)

nguyenquivinhquang

Data Preprocessing

1

Hi authors, thanks for the great work, are you planning to release the data preprocessing code? If yes, when will it be released?

botaoye

When Building wheel for detectron2 (setup.py), always fail

2

Hello. I ran into some problems, when I did "pip install -e .[all]".The system always prompts me, "Building wheel for detectron2 (setup.py) ... error error: subprocess-exited-with-error", and mentions "python setup.py...

XuanzheChen

DDP has slower training speed

3

Hi, I am trying to Finetune HMR2.0 on 3 RTX3090. It works fine when I only train it on 1 gpu. When I set trainer.devices=3, I received the error: ValueError:...

jyuntins

Meaning of keys in .pkl after running track.py

5

Hi! Thanks for your great work! As I finished running track.py and I got a .pkl file with following ```dict_keys(['time', 'shot', 'frame_path', 'tracked_ids', 'tracked_bbox', 'tid', 'bbox', 'tracked_time', 'appe', 'loca', 'pose',...

JerryPW

2d and 3d joints positions

2

I tried to use the smpl parameter to calculate the position of 3d joints in order to compare with the data under the 3d_joints key, but found that the two...

Marlod390

pose_transformer_v2 (BERT style transformer) use while running model on video

3

In the run tracking demo on videos, the paper mentions that the BERT-style transformer model(pose_transformer_v2) enables future predictions and amodal completion of missing detections within the same framework. However, in...

thribhuvanrapolu