search

karfly / learnable-triangulation-pytorch

This repository is an official PyTorch implementation of the paper "Learnable Triangulation of Human Pose" (ICCV 2019, oral). Proposed method archives state-of-the-art results in multi-view 3D human pose estimation!
MIT License
1.09k stars 181 forks source link

Please provide example of input #22

Closed exotikh3 closed 4 years ago

exotikh3 commented 4 years ago

Could you please provide example of input for model to be able to reproduce results. I requested access to Human3.6M but I think it would take a lot of time to receive approve.

karfly commented 4 years ago

@exotikh3

Hello! Unfortunately, right now I’m on the business trip and don’t have access to the workstation where data is stored. I’ll provide a minimal viable example as soon as I come back to work.

karfly commented 4 years ago

@exotikh3 @junhao1999, I dumped 1 big batch of Human3.6M, outputted right from dataloader here. You can download it from Google Drive.

Example of usage:

import pickle

with open("human36m_batch_128.pkl", 'rb) as fin:
    batch = pickle.load(fin)

...  # continue from line https://github.com/karfly/learnable-triangulation-pytorch/blob/master/train.py#L186
BingEdison commented 4 years ago

finally can try this awesomeness but getting errors from test output from the human3.6m batch provided, seems simple if an easy fix could you tell me what I'm doing wrong? Like others I do not have access to dataset so additional npy are missing

running eval code provided yields: python3 train.py \ --eval --eval_dataset val \ --config eval/human36m_vol_softmax.yaml \ --logdir ./logs

FileNotFoundError: [Errno 2] No such file or directory: './data/human36m/extra/human36m-multiview-labels-GTbboxes.npy'

karfly commented 4 years ago

Hi, @BingEdison. Please check step 3 at Human3.6M preprocessing: https://github.com/karfly/learnable-triangulation-pytorch/blob/master/mvn/datasets/human36m_preprocessing/README.md

BingEdison commented 4 years ago

Howdy @Karfly! however I do not have access to human3.6m dataset , I've downloaded the file 'bboxes-Human36M-GT' provided on google drive by you and team, following steps seem to require images and directories from human36m dataset is it still possible to complete instructions on above page you've referenced?

karfly commented 4 years ago

@BingEdison you can’t perform evaluation or training using dumped batch, I provided above. I shared it to allow people to explore how input data is organized without having whole Human3.6M dataset.

aryan-f commented 1 year ago

@exotikh3 @junhao1999, I dumped 1 big batch of Human3.6M, outputted right from dataloader here. You can download it from Google Drive.

Example of usage:

import pickle

with open("human36m_batch_128.pkl", 'rb) as fin:
    batch = pickle.load(fin)

...  # continue from line https://github.com/karfly/learnable-triangulation-pytorch/blob/master/train.py#L186

Could you share the batch again? The link doesn't seem to be working.

five1wheel commented 6 months ago

I'd like to request one, too