When training using `` - Githubissues

maudzung / TTNet-Real-time-Analysis-System-for-Table-Tennis-Pytorch

Unofficial implementation of "TTNet: Real-time temporal and spatial video analysis of table tennis" (CVPR 2020)

https://arxiv.org/pdf/2004.09927.pdf

586 stars 157 forks source link

When training using `` #2

Closed chenzhutian closed 4 years ago

chenzhutian commented 4 years ago

When I try to run the training using train_1st_phase, it shows that

fileNotFoundError: [Errno 2] No such file or directory: '../../dataset/training/images/game_2/img_111667.jpg'

Could you please help?

maudzung commented 4 years ago

Hi Chenzhutian,

How did you extract images from videos? You should execute python extract_smooth_labellings.py to extract images when you set an argument ----smooth-labelling in train bash shell scripts. (I will update the instruction for data preparation soon)

chenzhutian commented 4 years ago

Hi @maudzung ,

Thank you for your reply. I use python extract_selected_images.py to extract images from the videos. Now I am running the extract_all_images.py, but it is really slow.

maudzung commented 4 years ago

You can execute python extract_all_images.py to extract all images in videos, but I don't recommend to do that. To save your time and your resource, you just need to run python extract_smooth_labellings.py. Anw, grab a cup of coffee and have fun coding 👍

chenzhutian commented 4 years ago

Awesome. I have stoped and now runing extract_smooth_labellings.py. BTW, may I ask what are the differences between extract_smooth_labellings.py and extract_selected_images.py?

maudzung commented 4 years ago

extract_selected_images.py: Extract images based on events given in the annotation files. For example, if an event occurs at frame #10, we need to extract from frame #6 to frame #14.
extract_smooth_labellings.py: The smooth labeling technique is mentioned in the paper (part 5.2). This script is to extract images based on smoothed events.